From 6ade24f9ed79b0925a056caf55a964343755817a Mon Sep 17 00:00:00 2001 From: Uma Kumar Date: Tue, 1 Jul 2025 13:19:24 -0700 Subject: [PATCH 1/2] updating instructions for retail --- .../ai-vector-search-challenge-answers.md | 21 ++--- .../ai-exercise/ai-vector-search-challenge.md | 22 ++--- .../app-architecture/app-architecture.md | 4 +- dev-ai-app-dev-retail/build/build.md | 26 +++--- .../codingbasics/codingbasics.md | 6 +- .../connect-to-env/connect-to-env.md | 7 +- .../introduction/introduction.md | 8 +- .../user-story/user-story.md | 83 +++++++++---------- 8 files changed, 91 insertions(+), 86 deletions(-) diff --git a/dev-ai-app-dev-retail/ai-exercise/ai-vector-search-challenge-answers.md b/dev-ai-app-dev-retail/ai-exercise/ai-vector-search-challenge-answers.md index 1baacc9b..0db68a1d 100644 --- a/dev-ai-app-dev-retail/ai-exercise/ai-vector-search-challenge-answers.md +++ b/dev-ai-app-dev-retail/ai-exercise/ai-vector-search-challenge-answers.md @@ -4,15 +4,15 @@ **Welcome to the Code with AI Vector Search Challenge!** -In this lab, you will use Oracle Database 23ai to build a loan matching feature for a financial application. By implementing similarity search, you will match user queries to loan descriptions based on meaning—**not just keywords**. +In this lab, you upgrade a retail returns portal by adding a similar return search powered by AI Vector Search in Oracle Database 23ai. Working in a Jupyter Notebook, you’ll finish partially written Python that vectorizes product images, damage descriptions, and order notes so a new request can be matched to semantically similar past returns. The exercise shows how vector search spots fraud patterns and fast track approvals far better than basic keyword filters. ![Similarity Search](./images/similarity-search.png " ") -At SeerEquites, the data science team is building a loan recommendation system that goes beyond keyword matching. They are using AI Vector Search to understand meaning in loan descriptions—for example, linking "Affordable Housing Loan" to "Low-Income Home Loan." In this lab, you will complete their Similarity Search feature using Oracle Database 23ai, which converts text into embeddings and uses cosine similarity to find conceptually similar loans. +At Seer Retail, the data science team is building a return recommendation system that goes beyond keyword matching. They are using AI Vector Search to understand meaning in return descriptions—for example, linking "Damaged Return" to "Missing Return." In this lab, you will complete their Similarity Search feature using Oracle Database 23ai, which converts text into embeddings and uses cosine similarity to find conceptually similar returns. **Why AI Vector Search?** -Unlike keyword search, AI Vector Search captures meaning. For example, 'Military Veteran Loan' and 'Veteran Home Loan' use different words but express similar ideas. Oracle Database 23ai converts loan descriptions into vectors, compares their meanings, and ranks results by similarity for more accurate recommendations. +Unlike keyword search, AI Vector Search captures meaning. For example, 'Tampered Item' and 'Damaged Item' use different words but express similar ideas. Oracle Database 23ai converts return descriptions into vectors, compares their meanings, and ranks results by similarity for more accurate recommendations. Good luck, and enjoy the process. Let’s get coding! @@ -29,7 +29,7 @@ In this lab, you will: - Connect to Oracle Database 23ai and set up the environment. - Vectorize loan data from the `MOCK_LOAN_DATA` table into `LOAN_CHUNK`. - Verify the data was vectorized. -- Run a similarity search to find the top 5 loans matching a user’s query. +- Run a similarity search to find the top 5 returns matching a user’s query. ### Prerequisites: @@ -51,7 +51,7 @@ The data science team at SeerEquites has provided a Jupyter Notebook called `ai_ - **Connect to the database and import the necessary libraries and drivers.** - **Vectorize the `loan_type` column from the `MOCK_LOAN_DATA` table into the `LOAN_CHUNK` table.** - **Verify the vectorized data by displaying an output.** -- **Implement a Similarity Search to match a user's query to loans.** +- **Implement a Similarity Search to match a user's query to returns.** ## Task 2: Create a new Jupyter Notebook @@ -160,12 +160,12 @@ The second cell vectorizes loan descriptions from the `MOCK_LOAN_DATA` table, st ``` Vectorized 'loan_data' from 'MOCK_LOAN_DATA' and stored in 'LOAN_CHUNK'. ``` - - **This means the loan descriptions are now vectors, ready for similarity search.** + - **This means the return descriptions are now vectors, ready for similarity search.** ## Task 4: Verify the vectorized data -The third code block checks that the vectorization worked by querying the first three entries in LOAN_CHUNK. This lets you see the stored loan data, ensuring it’s ready for the next step. +The third code block checks that the vectorization worked by querying the first three entries in LOAN_CHUNK. This lets you see the stored return data, ensuring it’s ready for the next step. 1. Create a new cell and copy and paste the following code into it: @@ -280,7 +280,7 @@ The fourth code block ties it all together. It lets users enter a loan descripti - Joins **`LOAN_CHUNK`** and **`MOCK_LOAN_DATA`** to get loan details. - Filters for non-null vectors and CHUNK_ID >= 1000. - Orders the results by **`VECTOR_DISTANCE`** using cosine similarity to find the closest matches. - - Limits the results to display the top 5 loans. + - Limits the results to display the top 5 returns. ![Section of code that vectorizes the query](./images/fourth-cell-sql.png " ") @@ -320,12 +320,13 @@ The fourth code block ties it all together. It lets users enter a loan descripti (**Note**: Results may vary.) - - This shows the top 5 loans that are most similar in meaning to your query, even if the exact words differ. + - This shows the top 5 returns that are most similar in meaning to your query, even if the exact words differ. 4. Want to keep exploring? Try: - Entering different queries, like "Veteran Loan" or "Low-Income Housing," to see how results change. - Change Cosine to Euclidean, Dot, Manhattan or any of the other vector similarities and see how the results vary. +## Conclusion **Congratulations, you have successfully completed the AI Vector Search Challenge!** You have successfully: @@ -333,7 +334,7 @@ You have successfully: - Connected to Oracle Database 23ai. - Vectorized loan descriptions into embeddings. - Verified the stored data. -- Built a Similarity Search to match user queries to loans. +- Built a Similarity Search to match user queries to returns. By completing this challenge, you’ve seen how Oracle 23ai AI Vector Search transforms text into vectors and finds meaningful matches. diff --git a/dev-ai-app-dev-retail/ai-exercise/ai-vector-search-challenge.md b/dev-ai-app-dev-retail/ai-exercise/ai-vector-search-challenge.md index 0343da4f..6287f743 100644 --- a/dev-ai-app-dev-retail/ai-exercise/ai-vector-search-challenge.md +++ b/dev-ai-app-dev-retail/ai-exercise/ai-vector-search-challenge.md @@ -4,15 +4,15 @@ **Welcome to the Code with AI Vector Search Challenge!** -This lab guides you through building a loan matching feature for a financial app using Similarity Search in Oracle Database 23ai. You will fix and complete a Jupyter Notebook to implement AI Vector Search, enabling the app to find loan offers that closely match user profiles based on vector similarity. +In this lab, you upgrade a retail returns portal by adding a similar return search powered by AI Vector Search in Oracle Database 23ai. Working in a Jupyter Notebook, you’ll finish partially written Python that vectorizes product images, damage descriptions, and order notes so a new request can be matched to semantically similar past returns. The exercise shows how vector search spots fraud patterns and fast track approvals far better than basic keyword filters. ![Similarity Search](./images/similarity-search.png " ") -At SeerEquites, the data science team is building a smarter loan recommendation system using vector similarity. They started implementing a feature that vectorizes loan descriptions and matches them to user queries—**but the code is incomplete**. ❗Key sections are marked (**"CHANGES GO HERE"**) and it is up to you to finish the job and make the feature work.❗ +At Seer Retail, the data science team is building a smarter return recommendation system using vector similarity. They started implementing a feature that vectorizes return packages and matches them to user queries—**but the code is incomplete**. ❗Key sections are marked (**"CHANGES GO HERE"**) and it is up to you to finish the job and make the feature work.❗ **Why AI Vector Search?** -Keyword searches often miss context. For example, "Affordable Housing Loan" and "Low-Income Home Loan" mean similar things but use different terms. AI Vector Search solves this by converting text into vectors (embeddings) that capture meaning. With Oracle Database 23ai, you will use similarity search to match user queries to loan descriptions based on semantics, not just keywords. If you're ready to dive in, proceed with this lab and start coding. If you prefer a more guided approach with a detailed walkthrough, you can continue to Lab 5b Step-by-step: Code with AI Vector Search for step-by-step instructions. +Keyword searches often miss context. For example, "Damaged Return" and "Missing Returns" mean similar things but use different terms. AI Vector Search solves this by converting text into vectors (embeddings) that capture meaning. With Oracle Database 23ai, you will use similarity search to match user queries to loan descriptions based on semantics, not just keywords. If you're ready to dive in, proceed with this lab and start coding. If you prefer a more guided approach with a detailed walkthrough, you can continue to Lab 5b Step-by-step: Code with AI Vector Search for step-by-step instructions. Good luck, and enjoy the process. Let’s get coding! @@ -28,7 +28,7 @@ In this lab, you will: - Vectorize loan data from the `MOCK_LOAN_DATA` table into `LOAN_CHUNK` table. - Verify the data was vectorized. -- Implement a similarity search to find the top 5 matching loans based on a user’s query. +- Implement a similarity search to find the top 5 matching returns based on a user’s query. ### Prerequisites: @@ -45,7 +45,7 @@ This lab assumes you have: ## Task 1: Challenge requirements -The data science team at SeerEquites has provided a Jupyter notebook called `ai_vector_search_challenge.ipynb` with four key code cells. The first cell sets up the environment, but the next three cells which are responsible for vectorizing data, verifiying it and performing similarity search contain errors marked as ❗**"CHANGES GO HERE"**❗. Your job is to fix these cells in order to make the feature work. The notebook will: +The data science team at Seer Retail has provided a Jupyter notebook called `ai_vector_search_challenge.ipynb` with four key code cells. The first cell sets up the environment, but the next three cells which are responsible for vectorizing data, verifiying it and performing similarity search contain errors marked as ❗**"CHANGES GO HERE"**❗. Your job is to fix these cells in order to make the feature work. The notebook will: - **Connect to the database and import the necessary libraries and drivers.** - **Vectorize the `loan_type` column from the `MOCK_LOAN_DATA` table into the `LOAN_CHUNK` table.** @@ -140,7 +140,7 @@ However, the SQL `INSERT` statement has missing code where the vector is generat ``` -2. ❗Look for the lines with **(CHANGES GO HERE)** inside the `INSERT statement:❗ +2. ❗Look for the lines with **(CHANGES GO HERE)** inside the INSERT statement:❗ ![Second code cell](./images/vectorize-loan-data.png " ") @@ -209,7 +209,7 @@ The third function verifies vectorization by querying the `LOAN_CHUNK` table for ## Task 6: Implement similarity search -The fourth code block is the core of the challenge: letting users enter a loan description and retrieve the top 5 most similar loans. The function looking\_for_similar\_loans(query) should: +The fourth code block is the core of the challenge: letting users enter a return description and retrieve the top 5 most similar returns. The function looking\_for_similar\_loans(query) should: - Convert the user’s input into a vector. @@ -339,16 +339,18 @@ The fourth code block is the core of the challenge: letting users enter a loan d (**Note:** Results may vary.) +## Conclusion + **Congratulations, you have successfully completed the AI Vector Search Challenge!** You have successfully: - Connected to Oracle Database 23ai. -- Vectorized loan descriptions into embeddings. +- Vectorized return descriptions into embeddings. - Verified the stored data. -- Built a Similarity Search to match user queries to loans. +- Built a Similarity Search to match user queries to returns. -Fixing the code allowed SeerEquites to recommend loans based on meaning, not keywords, using AI Vector Search in Oracle Database 23ai. You learned to generate embeddings, query vectors, and integrate AI with SQL—skills you can apply to other projects. +Fixing the code allowed Seer Retail to recommend return based on meaning, not keywords, using AI Vector Search in Oracle Database 23ai. You learned to generate embeddings, query vectors, and integrate AI with SQL—skills you can apply to other projects. ## Learn More diff --git a/dev-ai-app-dev-retail/app-architecture/app-architecture.md b/dev-ai-app-dev-retail/app-architecture/app-architecture.md index 0ac5630a..895e7025 100644 --- a/dev-ai-app-dev-retail/app-architecture/app-architecture.md +++ b/dev-ai-app-dev-retail/app-architecture/app-architecture.md @@ -8,7 +8,7 @@ Estimated Lab Time: 15 minutes ## Physical Architecture -The SeerEquities loan application runs in an **Oracle Cloud Infrastructure (OCI)** Region, with its application layer in a public subnet inside a **Virtual Cloud Network (VCN)**. +The Seer Retail return authorization application runs in an **Oracle Cloud Infrastructure (OCI)** Region, with its application layer in a public subnet inside a **Virtual Cloud Network (VCN)**. ![Physical Architecture](./images/physical-architecture.png) @@ -63,7 +63,7 @@ Oracle Database 23ai supports property graphs, which model relationships using v Property graphs simplify working with connected data, such as identifying influencers, predicting trends, or discovering relationships through pattern matching, by offering a more intuitive and efficient way to model and query complex networks. -**Where is it used**: We implemented property graphs in the demo part of this workshop. Loan approval officers can use it to identify potential alternative loans for their customers. +**Where is it used**: We implemented property graphs in the demo part of this workshop. Return Authorization Specialists can use it to identify potential return recommendations for their customers. ## Acknowledgements diff --git a/dev-ai-app-dev-retail/build/build.md b/dev-ai-app-dev-retail/build/build.md index 70c81da1..2e54ce92 100644 --- a/dev-ai-app-dev-retail/build/build.md +++ b/dev-ai-app-dev-retail/build/build.md @@ -2,7 +2,7 @@ ## Introduction -In this lab, you will build a loan recommendation system powered by Oracle Database 23ai and OCI Generative AI. You will connect to a Oracle Database 23ai, explore customer data, extract relevant insights, and use a Large Language Model (LLM) to generate personalized loan suggestions. The system combines AI with OCI Generative AI to provide personalized loan recommendations. This system integrates AI capabilities with Oracle's robust database technology to deliver targeted financial recommendations. +In this lab, you build a complete return recommendation engine with Oracle Database 23ai and OCI Generative AI. Connect to the database, explore order and image data, and invoke a large language model to generate personalized return decisions and policy explanations. Building on earlier exercises, you’ll apply Python to deliver a fully integrated, AI-powered retail returns application. This lab uses some of the basic coding samples you created in lab 3, such as cursor.execute and more. @@ -15,8 +15,8 @@ To get things started we invite you to watch this video and see the lab in actio ### Objectives -* Build the complete loan approval application as seen in lab 1 -* Use OCI Generative AI to generate personalized loan recommendations +* Build the complete return authorization application as seen in lab 1 +* Use OCI Generative AI to generate personalized risk score and return recommendations * Use Python to connect to an Oracle Database 23ai instance and run queries * Explore customer data and extract relevant information @@ -139,25 +139,25 @@ With the database ready, you will query customer data from the clients_dv JSON d ![Fetch customer](./images/fetch-customer.png " ") -If you completed Lab 1: Run the Demo earlier, this is what gets printed out when the Loan Officer clicks on CUST 1000. You just built it, well done! +If you completed Lab 1: Run the Demo earlier, this is what gets printed out when the Return Analyst clicks on CUST 1000. You just built it, well done! ## Task 5: Create a function to generate recommendations for the customer -With customer profiles in place, you will use OCI Generative AI to generate personalized loan recommendations. This step combines customer data with available loan options, allowing the LLM to suggest loans that match the customer’s credit score, income, and debt profile. +With customer profiles in place, you will use OCI Generative AI to generate personalized return recommendations. This step combines historical return records with customer data, allowing the LLM to suggest return recommendations that match the customer’s return frequency, past behavior, and lifetime value. Here’s what we’ll do: -- **Fetch Loan Options**: Retrieve all loans from `MOCK_LOAN_DATA`, including details like interest rates, credit score requirements, and loan types. +- **Fetch Return Options**: Retrieve all returns from `MOCK_LOAN_DATA`, including details like product category, price tier, and damage descriptors. - **Build a Prompt**: Construct a structured prompt that combines the customer’s profile with available loans, instructing the LLM to evaluate and recommend based solely on this data. - **Use OCI Generative AI**: Send the prompt to the `meta.llama-3.2-90b-vision-instruct` model via OCI’s inference client, which will process the input and generate a response. - **Format the Output**: Display the recommendations in HTML with styled headers and lists, covering evaluation, top picks, and explanations—making it easy to read and understand. -1. Review and review the code in a new cell: +1. Run and review the code in a new cell: ```python - # Fetch Mock Loan Data + # Fetch Mock Return Data cursor.execute("SELECT loan_id, loan_provider_name, loan_type, interest_rate, origination_fee, time_to_close, credit_score, debt_to_income_ratio, income, down_payment_percent, is_first_time_home_buyer FROM MOCK_LOAN_DATA") df_mock_loans = pd.DataFrame(cursor.fetchall(), columns=["LOAN_ID", "LOAN_PROVIDER_NAME", "LOAN_TYPE", "INTEREST_RATE", "ORIGINATION_FEE", "TIME_TO_CLOSE", "CREDIT_SCORE", "DEBT_TO_INCOME_RATIO", "INCOME", "DOWN_PAYMENT_PERCENT", "IS_FIRST_TIME_HOME_BUYER"]) @@ -185,7 +185,7 @@ Here’s what we’ll do: print(recommendations) ``` -2. Click the "Run" button to execute the code. Note that this will take time to run. Be patient, you will get loan recommendations from the LLM shortly. +2. Click the "Run" button to execute the code. Note that this will take time to run. Be patient, you will get return recommendations from the LLM shortly. 3. Review the output. In the demo, this is where you selected the "Navigate to Decisions" button as the Loan Officer. You just used AI to get recommendations for the loan officer which would have taken her hours to do, congratulations! @@ -196,9 +196,9 @@ Here’s what we’ll do: ## Task 6: Create a function to create embeddings - Use Oracle Database 23ai's to create vector data -To handle follow-up questions, you will enhance the system with an AI Loan Guru powered by Oracle 23ai’s Vector Search and Retrieval-Augmented Generation (RAG). The AI Loan Guru will be able to answer questions about the loan application and provide recommendations based on the data. +To handle follow-up questions, you will enhance the system with a virtual return officer assistant, Riley, powered by Oracle 23ai’s Vector Search and Retrieval-Augmented Generation (RAG). Riley will be able to answer questions about the return authorization and provide recommendations based on the data. -Before answering questions, we need to prepare the data by vectoring the loan recommendations. This step: +Before answering questions, we need to prepare the data by vectoring the return recommendations. This step: - **Stores Recommendations**: Inserts the full recommendation text (from previous cell) as a single chunk if not already present. @@ -248,7 +248,7 @@ Before answering questions, we need to prepare the data by vectoring the loan re Now that the recommendations are vectorized, we can process a user’s question: - ```What’s the best loan for a first-time home buyer?``` + ```What is the risk score for this buyer?``` This step: @@ -400,7 +400,7 @@ Now that the recommendations are vectorized, we can process a user’s question: ![rag](./images/rag.png " ") -## Summary +## Conclusion Congratulations! You implemented a RAG process in Oracle Database 23ai using Python. diff --git a/dev-ai-app-dev-retail/codingbasics/codingbasics.md b/dev-ai-app-dev-retail/codingbasics/codingbasics.md index 32f4d923..ab971068 100644 --- a/dev-ai-app-dev-retail/codingbasics/codingbasics.md +++ b/dev-ai-app-dev-retail/codingbasics/codingbasics.md @@ -2,7 +2,7 @@ ## Introduction -In this hands-on lab, you will learn the basics of working with Oracle Database 23ai using Python. You will connect to the database with both `oracledb` and `pymongo`, create and query tables, and build JSON Duality Views. You will also explore Oracle’s Mongo API to query and update data using a familiar document model. For organizations already using MongoDB-like applications, the Oracle MongoDB API enables a smoother transition and integration with existing code, tools, or drivers—while consolidating everything in Oracle’s converged platform. +In this hands-on lab, you'll use Python to interact with Oracle Database 23ai - connecting with both `oracledb` and `pymongo`, creating tables, querying data, and building JSON Duality Views. You'll also explore Oracle’s Mongo API to query and update data using a document-based approach, making it easy to integrate with existing MongoDB-style apps. This lab is ideal for those with basic Python and Oracle knowledge and builds on the setup from Lab 2. Estimated Time: 30 minutes @@ -419,7 +419,7 @@ Next, let's update some data in our database using MongoDB syntax. Let's write a The final step in our basic coding tour with Python and the Oracle Database 23ai is to query the two relational tables we created earlier. We just updated the JSON Duality view, so let's run a query on it to see if the changes were reflected. Remember that we created two functions in the beginning of this lab: `query_customers()` and `query_orders()`. Let's use these functions to get our data from MongoDB and update it with the new email address. -1. First, run the ``query_customers()` function in a new cell +1. First, run the `query_customers()` function in a new cell ```python @@ -478,7 +478,7 @@ The final step in our basic coding tour with Python and the Oracle Database 23ai 🔴 **`df.head()`** - This is a method that returns that returns the result of the query including the column names. -## Summary +## Conclusion In this lab, we learned how to use Python and Oracle's Python driver `oracledb` to interact with Oracle Database 23ai's new features. You learned how to user the `cursor` object to execute SQL queries. Using the `cursor` object, you created a **JSON Duality View** and you even used some JSON functions to query documents using SQL syntax. Then, you also learned how to connect to the database using `pymongo` and retrieve data from a table in the database using **MongoDB syntax**. You created functions to update the **JSON Duality View** and you learned how these updates are also reflected in the underlying relational database tables. diff --git a/dev-ai-app-dev-retail/connect-to-env/connect-to-env.md b/dev-ai-app-dev-retail/connect-to-env/connect-to-env.md index 3a0676a4..3e5f4698 100644 --- a/dev-ai-app-dev-retail/connect-to-env/connect-to-env.md +++ b/dev-ai-app-dev-retail/connect-to-env/connect-to-env.md @@ -2,7 +2,7 @@ ## Introduction -In this lab, we will show you how to connect to the development environment. We are using JupyterLab as our development environment, which is a web-based application that allows you to write and run code on your browser. You can use this environment to create notebooks, edit files, and run Python scripts. +In this lab, you'll learn how to access the web-based Jupyter Lab development environment, where you can write and run Python code directly in your browser. You'll log in, create notebooks, and get ready to start coding in a hands-on environment. Estimated Time: 5 minutes @@ -36,7 +36,7 @@ To get things started we invite you to watch this video and see the lab in actio ## Task 3: Get familiar with Jupyter Notebooks -You will use a Jupyter Notebook in JupyterLab to build and test the loan recommendation system. If you are new to notebooks, the following tips will help you get started and work smoothly. +You will use a Jupyter Notebook in JupyterLab to build and test the return authorization system. If you are new to notebooks, the following tips will help you get started and work smoothly. 1. **Executing Code Blocks**: You can run code in two simple ways: press **Shift+Enter** to execute and move to the next cell, or click the **Play/Execute** button in the menu bar at the top of this tab. Both methods work interchangeably. @@ -48,6 +48,9 @@ You will use a Jupyter Notebook in JupyterLab to build and test the loan recomme ![JupyterLab blocks](./images/block.png " ") +## Conclusion + +In this lab you logged into the **IDE Development Environment** for Jupyter Labs. You reviewed different elements from the environment like **File Browser** and **Launcher**. Lastly, you were introduced to concepts to allow you to get started running code smoothly like **executing code blocks** and **outputs and warnings**. ## Acknowledgements diff --git a/dev-ai-app-dev-retail/introduction/introduction.md b/dev-ai-app-dev-retail/introduction/introduction.md index aba0cb87..fcb71f1a 100644 --- a/dev-ai-app-dev-retail/introduction/introduction.md +++ b/dev-ai-app-dev-retail/introduction/introduction.md @@ -6,7 +6,7 @@ Generative AI and Oracle Database 23ai makes it possible to unlock deeper insights, faster—directly within your database. -In this hands-on workshop, you’ll build a finance-focused GenAI solution using Oracle Database 23ai. +In this hands-on workshop, you’ll build a retail-focused GenAI solution using Oracle Database 23ai. To get things started we invite you to watch the introduction video to see the demo in action: @@ -17,7 +17,7 @@ You will: * Combine your **enterprise data with large language models (LLMs)** using Retrieval-Augmented Generation (RAG) to answer natural language questions with context-aware accuracy. - * Automate real-world financial workflows, like loan approvals, by **embedding and querying unstructured data** in real time. + * Automate real-world retail workflows, like return authorizations, by **embedding and querying unstructured data** in real time. By the end of this workshop, you’ll understand how Oracle Database 23ai helps financial organizations: @@ -57,11 +57,11 @@ In lab 4, you will learn how to **implement a working example of retrieval-augme #### ➡️ **Start experimenting!** -You can also experiment with the code in your own environment. The workshop environment is flexible and allows you try out different things: Start building your own version of the loan approval application! +You can also experiment with the code in your own environment. The workshop environment is flexible and allows you try out different things: Start building your own version of the return authorization application! ### Objectives -* Learn how Oracle Database 23ai capabilities can be used to automate and enhance the loan approval process +* Learn how Oracle Database 23ai capabilities can be used to automate and enhance the return authorization process * Start coding with Python and Oracle's Python driver `oracledb` diff --git a/dev-ai-app-dev-retail/user-story/user-story.md b/dev-ai-app-dev-retail/user-story/user-story.md index c70267d1..59c3652f 100644 --- a/dev-ai-app-dev-retail/user-story/user-story.md +++ b/dev-ai-app-dev-retail/user-story/user-story.md @@ -2,7 +2,7 @@ ## Introduction -In this lab, you will step into the role of a loan officer using a next-generation loan approval application powered by Oracle Database 23ai. You will work with real loan applications and see how Generative AI, Vector Search, and Graph analytics replace manual review with faster, AI-driven decision-making. +In this lab, you’ll step into the shoes of a **Return Authorization Specialist** using an AI powered returns app built on Oracle Database 23ai. See how Generative AI, Vector Search, and Graph Analytics reduce review times by automating photo analysis, order verification, and fraud checks - all inside one converged database, with zero data shuffling. **Disclaimer**: Please note that your results may vary. The information provided is generated by OCI Generative AI services, and your outcomes may differ from those presented. @@ -12,7 +12,7 @@ Estimated Lab Time: 30 minutes In this lab, you will: -* Review how the Seer Equities Loan approval app incorporates the use of JSON Duality Views, Graph analytics, and other converged database features, all without requiring complex data movement or separate systems. +* Review how the Seer Retail return authorization app incorporates the use of AI Vector Search, Graph analytics, and other converged database features, all without requiring complex data movement or separate systems. ### Prerequisites @@ -30,30 +30,29 @@ This lab assumes you have: ![Login](./images/login.png =50%x*) -3. Welcome to the SeerEquities Loan Management application! Congratulations, you are now connected to the demo environment. You can now execute the different tasks for this Lab. +3. Welcome to the Seer Retail return authorization application! Congratulations, you are now connected to the demo environment. You can now execute the different tasks for this Lab. ![Homepage](./images/app-home.png =50%x*) -## Task 2: Demo - Customer with strong credit score +## Task 2: Demo - Customer with low risk score -In this first example, you will use the application to approve a customer with strong credit score. The first user on your to-do list is James Smith. +In this first example, you will use the application to authorize a return for a customer with a low risk score. The first user on your to-do list is James Smith. 1. On the Dashboard page, from the pending review list, select the Customer ID for **James Smith**. ![Select James Smith](./images/james-smith.png =50%x*) -2. Opening James Smith’s profile reveals his loan application details—name, location, requested amount, debt, and credit score. +2. Opening James Smith’s profile reveals his return authorization details—name, location, requested amount, debt, and credit score. ![James Smith AI generated recommendations](./images/james-smith-ai.png =50%x*) -3. At the bottom of James Smith’s profile, you will find the **AI Loan Guru**—a chatbot built on Oracle Database 23ai and Vector search. When prompted, the system uses **RAG** to generate a response. It converts the question and loan data into embeddings, performs a similarity search, and then uses the **GenAI service** to turn the enriched context into a clear, natural language answer. If the customer calls with a question, you can quickly enter it into the AI Loan Guru to generate a relevant response. - +3. At the bottom of James Smith’s profile, you will find **AI Return Specialist**—a virtual Return‑Authorization Specialist who guides attendees through every step of the AI‑driven returns portal. Riley shows how to upload photos and order data, highlights how AI Vector Search compares each request to millions of past returns, and translates risk scores into plain language. Her friendly tips humanize concepts like graph‑based customer‑lifetime‑value analysis and RAG‑generated policy explanations, so users of any technical level can confidently approve, pend, or deny a return - **Copy** the question below into the AI chatbot and press **Enter**. What does the AI Loan Guru recommend? + **Copy** the question below into the AI chatbot and press **Enter**. What does AI Return Specialist say? ```text - What would be the top 0% down payment loans? + Has this customer returned similar items before? ``` @@ -65,21 +64,21 @@ In this first example, you will use the application to approve a customer with s ![James Smith Decision](./images/james-smith-decision.png =50%x*) - After navigating to the decisions page, the AI evaluation runs in the background. It analyzes James’s profile and matches it against available loan options in the database. A custom AI prompt ensures the system uses only internal data—never the internet. In this case, the AI returns three loan options, each with a clear explanation. All options are displayed alongside the AI’s final recommendation: approval. + After navigating to the decisions page, the AI evaluation runs in the background. Using AI Vector Search, it semantically matches the content against thousands of historical return records—identifying patterns that suggest either legitimate claims or potential abuse. At the same time, Graph Analytics examines the customer’s prior behavior, highlighting unusual return frequency or links to high-risk product categories. For each case, the system assigns a risk score ranging from Low to Very High and generates a clear recommendation—such as “Approved – consistent with customer profile” or “Request More Info – insufficient documentation.” -5. In the **Select Final Approved Loan Option** section, the available loan options are displayed. The loan officer can choose to request more information to refine the offer, but for this scenario, we will proceed by selecting one of the suggested loans. +5. In the **Select Final Approved Return Option** section, the risk score and decision are displayed. Once a decision is made, the analyst finalizes the return. A professionally formatted PDF letter is automatically generated, summarizing the outcome and rationale. The customer receives an email moments later, clearly stating whether their return was approved, denied, or needs more information—along with an explanation grounded in policy and precedent. ![James Smith Select Final Approved Loan](./images/james-smith-select-final-loan.png =50%x*) -6. Select the AI-recommended loan labeled as the military veteran loan with the shortest time to close. In this example, that is Loan 26—the only military veteran option shown. +6. Select the AI-recommended return decision. In this example, that is **Approved**. >Please note that your results may vary. The information provided is generated by OCI Generative AI services, and your outcomes may differ from those presented. ![James Smith Select Loan 3](./images/james-smith-select-loan.png =50%x*) -7. Set the final loan status to **Approved**, then click **Save Final Approval & Loan Status** to complete the process. +7. Set the final return status to **Approved**, then click **Save Final Approval & Return Status** to complete the process. - The loan status has been updated to 'Approved' and saved to the customer profile. + The return status has been updated to 'Approved' and saved to the customer profile. ![James Smith Save Loan Status](./images/james-smith-save.png =50%x*) @@ -99,21 +98,21 @@ In this first example, you will use the application to approve a customer with s ![Return to Dashboard](./images/james-return-dashboard.png =50%x*) -12. Expand **View Approved Customers**. We can see that James Smith has been removed from the Pending Review Loan Requests list and has been added to the Approved Loan Requests list. +12. Expand **View Approved Customers**. We can see that James Smith has been removed from the Pending Review Return Requests list and has been added to the Approved Return Requests list. ![James Smith Approved List](./images/james-smith-approved-list.png =50%x*) **Conclusion** -Once you select and save one of the 3 loans recommended by the AI: +Once you select and save one of the 3 return options recommended by the AI: -✅ The customer's loan status is updated. +✅ The customer's return status is updated. ✅ A finalized PDF decision document is generated. ✅ The dashboard reflects the change in real-time — marking James as Approved. -Congratulations, you have just approved your first customer loan! Proceed to the next task. +Congratulations, you have just approved your first customer return! Proceed to the next task. ## Task 3: Demo - Denying a customer with high risk In this example, you will navigate the application to review a customer and deny them as part of the exercise. The second user on your to-do list is James Woods. @@ -122,13 +121,13 @@ In this example, you will navigate the application to review a customer and deny ![Select James Woods](./images/james-woods.png =50%x*) -2. Opening James Woods’s profile displays his loan application details. Within a few seconds, the AI generates automatically recommendations. In this case, the system evaluates a less favorable profile and identifies key risk factors. +2. Opening James Woods’s profile displays his return package details. Within a few seconds, the AI generates automatically recommendations. In this case, the system evaluates a less favorable profile and identifies key risk factors. - This customer has: + This return has: [placeholder] - * **$5000 in income** - * A **low credit score** - * And is requesting a **high loan amount** + * **Identifiable Damage** + * An **unusual return frequency** + * And has links to **high risk product categories** Despite the risk factors, the AI evaluates the profile and suggests next steps. In this case, it recommends a denial—but also provides clear, actionable guidance to help the customer improve their chances of approval in the future. @@ -138,26 +137,26 @@ In this example, you will navigate the application to review a customer and deny ![James Woods Decision](./images/james-woods-decision.png =50%x*) ->⁉️ **What are two reasons that the AI decided to deny this applicant?** ⁉️ +>⁉️ **What are two reasons that the AI decided to deny this return?** ⁉️ -4. Expand **View Dot Plot: Client-Loan Relationships** to view the graph. +4. Expand **View Dot Plot: Customer-Return Relationships** to view the graph. ![James Woods Graph](./images/expand-graph.png =50%x*) - On the decision page, the loan officer can use **Operational Property Graph** to explore near-approval loan scenarios. Built with **Oracle Graph**, this feature visually maps loans just out of reach and highlights what adjustments—like increasing income or reducing debt—could improve eligibility. + On the decision page, the return specialist can use **Operational Property Graph** to explore near-approval return scenarios. Built with **Oracle Graph**, this feature visually maps loans just out of reach and highlights what adjustments—like increasing loyalty tier or reducing refund frequency—could improve eligibility. ![James Woods Graph](./images/james-woods-graph2.png =50%x*) >💡 In Oracle Database 23ai, **Property Graph** allows you to treat your data like a network of connected points, where each point (called a node) and each link (called an edge) has its own details or properties. This setup helps you run graph analytics, like finding important connections or patterns, directly within the database. -5. The loan status is set to **Denied**. Click the **Save Final Approval & Loan Status** button. +5. The return status is set to **Denied**. Click the **Save Final Approval & Return Status** button. - The loan status has been updated to 'Denied' and saved to the customer profile. + The return status has been updated to 'Denied' and saved to the customer profile. ![James Woods Save Loan Status](./images/james-woods-save.png =50%x*) -6. Press the **Generate Decision as PDF** button to save the AI responses and proceed to the final loan disposition. +6. Press the **Generate Decision as PDF** button to save the AI responses and proceed to the final return authorization. ![James Woods PDF](./images/james-woods-pdf.png =50%x*) @@ -173,23 +172,23 @@ In this example, you will navigate the application to review a customer and deny ![Return to Dashboard](./images/woods-return-dashboard.png =50%x*) -10. Expand **View Denied Customers**. You will see that James Woods has been moved from the **Pending Review Loan Requests** list to the **Denied Loan Requests** list. +10. Expand **View Denied Customers**. You will see that James Woods has been moved from the **Pending Review Return Requests** list to the **Denied Return Requests** list. ![James Woods Denied List](./images/james-woods-denied-list.png =50%x*) **Conclusion** -Congratulations, you have finished reviewing a customer with high financial risk! Proceed to the next task. +Congratulations, you have finished reviewing a customer with high fraud risk! Proceed to the next task. ## Task 4: Demo - Update customer details -Lastly, let’s explore how the system uses JSON Duality Views to handle profile updates. In this task, you will edit a customer’s details. In this example, the customer was asked to submit updated proof of income. +Lastly, let’s explore how the system uses JSON Duality Views to handle profile updates. In this task, you will edit a customer’s details. In this example, the customer was asked to submit updated proof of carrier tracking. -1. On the Dashboard page, from the **Pending Review Loan Requests** list, select the customer ID for **Alex Anderson**. +1. On the Dashboard page, from the **Pending Review Return Requests** list, select the customer ID for **Alex Anderson**. ![Select Alex Anderson](./images/select-alex-anderson.png =50%x*) -2. We will upload a document to update Alex's income. Before uploading the document, note that the customer’s income is currently listed as $25,000. On the Customer Details page, click the **Upload Document** button. +2. We will upload a document to update the item's condition. Before uploading the document, note that the item's condition is currently listed as "none". On the Customer Details page, click the **Upload Document** button. ![Upload Document](./images/upload-document.png =50%x*) @@ -203,7 +202,7 @@ Lastly, let’s explore how the system uses JSON Duality Views to handle profile ![Customer Profile Updated](./images/profile-updated.png) -5. Refresh the page and note that the income has been updated to $65,000. Thanks to JSON Transform and JSON Duality Views, only the relevant field is modified — leaving the rest of the profile UNTOUCHED. +5. Refresh the page and note that the item condition has been updated to "Damaged". Thanks to JSON Transform and JSON Duality Views, only the relevant field is modified — leaving the rest of the profile UNTOUCHED. ![Customer Profile Updated](./images/income-updated.png) @@ -211,23 +210,23 @@ Lastly, let’s explore how the system uses JSON Duality Views to handle profile Once the document is uploaded: -✅ The system automatically detects the new income data. +✅ The system automatically detects the new item condition data. -✅ Then their profile will be updated from $25,000 to $65,000. +✅ Then their profile will be updated from "none" to "Damaged". ✅ And thanks to JSON Transform and JSON Duality Views, only the relevant field is modified — leaving the rest of the profile UNTOUCHED. -## Summary +## Conclusion -In conclusion our Loan Management App was able to leverage Oracle database 23ai technologies such as **AI Vector Search, Property Graph and JSON Duality Views** to: +In conclusion our Return Authorization App was able to leverage Oracle database 23ai technologies such as **AI Vector Search, Property Graph and JSON Duality Views** to: ✅ Automate profile evaluations -✅ Provide AI-driven loan recommendations by using an RAG model powered by a Oracle Database 23ai's AI Vector Search and OCI Generative AI service +✅ Provide AI-driven return recommendations by using an RAG model powered by a Oracle Database 23ai's AI Vector Search and OCI Generative AI service ✅ Enable seamless profile updates with JSON Duality Views -✅ And empower loan officers with actionable insights through Operational Property Graphs +✅ And empower return analysts with actionable insights through Operational Property Graphs By combining these advanced tools, the application enables faster, smarter decisions and delivers clear guidance on how customers can improve their eligibility. From 4226cb367b3d4e94b17ce13a3e0a505017555a95 Mon Sep 17 00:00:00 2001 From: Uma Kumar Date: Mon, 7 Jul 2025 14:16:39 -0700 Subject: [PATCH 2/2] updates to healthcare markdown --- .../ai-vector-search-challenge-answers.md | 10 +- .../ai-exercise/ai-vector-search-challenge.md | 10 +- .../app-architecture/app-architecture.md | 4 +- dev-ai-app-dev-healthcare/build/build.md | 8 +- .../cloud-login/cloud-login.md | 6 +- .../codingbasics/codingbasics.md | 11 +- .../connect-to-env/connect-to-env.md | 8 +- .../introduction/introduction.md | 28 +++-- .../user-story/user-story.md | 100 +++++++++--------- .../workshops/sandbox/manifest.json | 2 +- .../workshops/tenancy/manifest.json | 2 +- .../workshops/sandbox/manifest.json | 2 +- .../workshops/tenancy/manifest.json | 2 +- 13 files changed, 104 insertions(+), 89 deletions(-) diff --git a/dev-ai-app-dev-healthcare/ai-exercise/ai-vector-search-challenge-answers.md b/dev-ai-app-dev-healthcare/ai-exercise/ai-vector-search-challenge-answers.md index 1baacc9b..0584dec7 100644 --- a/dev-ai-app-dev-healthcare/ai-exercise/ai-vector-search-challenge-answers.md +++ b/dev-ai-app-dev-healthcare/ai-exercise/ai-vector-search-challenge-answers.md @@ -4,15 +4,15 @@ **Welcome to the Code with AI Vector Search Challenge!** -In this lab, you will use Oracle Database 23ai to build a loan matching feature for a financial application. By implementing similarity search, you will match user queries to loan descriptions based on meaning—**not just keywords**. +Enhance your app with an intent-matching feature using AI Vector Search. Work in a Jupyter Notebook to enable similarity search on patient history, doctor notes, and past approvals. Match cases semantically—not just by keywords—and return relevant prior examples to support real-time decisions. ![Similarity Search](./images/similarity-search.png " ") -At SeerEquites, the data science team is building a loan recommendation system that goes beyond keyword matching. They are using AI Vector Search to understand meaning in loan descriptions—for example, linking "Affordable Housing Loan" to "Low-Income Home Loan." In this lab, you will complete their Similarity Search feature using Oracle Database 23ai, which converts text into embeddings and uses cosine similarity to find conceptually similar loans. +At Seer Healthcare, the data science team is building a claim recommendation system that goes beyond keyword matching. They are using AI Vector Search to understand meaning in treatment descriptions—for example, linking "Patient History" to "Prior Claims." In this lab, you will complete their Similarity Search feature using Oracle Database 23ai, which converts text into embeddings and uses cosine similarity to find conceptually similar claims. **Why AI Vector Search?** -Unlike keyword search, AI Vector Search captures meaning. For example, 'Military Veteran Loan' and 'Veteran Home Loan' use different words but express similar ideas. Oracle Database 23ai converts loan descriptions into vectors, compares their meanings, and ranks results by similarity for more accurate recommendations. +Unlike keyword search, AI Vector Search captures meaning. For example, '[placeholder]]' and '[placeholder]' use different words but express similar ideas. Oracle Database 23ai converts claim descriptions into vectors, compares their meanings, and ranks results by similarity for more accurate recommendations. Good luck, and enjoy the process. Let’s get coding! @@ -331,9 +331,9 @@ The fourth code block ties it all together. It lets users enter a loan descripti You have successfully: - Connected to Oracle Database 23ai. -- Vectorized loan descriptions into embeddings. +- Vectorized treatment descriptions into embeddings. - Verified the stored data. -- Built a Similarity Search to match user queries to loans. +- Built a Similarity Search to match user queries to treatments. By completing this challenge, you’ve seen how Oracle 23ai AI Vector Search transforms text into vectors and finds meaningful matches. diff --git a/dev-ai-app-dev-healthcare/ai-exercise/ai-vector-search-challenge.md b/dev-ai-app-dev-healthcare/ai-exercise/ai-vector-search-challenge.md index 0343da4f..32818a44 100644 --- a/dev-ai-app-dev-healthcare/ai-exercise/ai-vector-search-challenge.md +++ b/dev-ai-app-dev-healthcare/ai-exercise/ai-vector-search-challenge.md @@ -4,15 +4,15 @@ **Welcome to the Code with AI Vector Search Challenge!** -This lab guides you through building a loan matching feature for a financial app using Similarity Search in Oracle Database 23ai. You will fix and complete a Jupyter Notebook to implement AI Vector Search, enabling the app to find loan offers that closely match user profiles based on vector similarity. +Enhance your app with an intent-matching feature using AI Vector Search. Work in a Jupyter Notebook to enable similarity search on patient history, doctor notes, and past approvals. Match cases semantically—not just by keywords—and return relevant prior examples to support real-time decisions. ![Similarity Search](./images/similarity-search.png " ") -At SeerEquites, the data science team is building a smarter loan recommendation system using vector similarity. They started implementing a feature that vectorizes loan descriptions and matches them to user queries—**but the code is incomplete**. ❗Key sections are marked (**"CHANGES GO HERE"**) and it is up to you to finish the job and make the feature work.❗ +At Seer Healthcare, the data science team is building a claim recommendation system that goes beyond keyword matching. They are using AI Vector Search to understand meaning in treatment descriptions—for example, linking "Patient History" to "Prior Claims." In this lab, you will complete their Similarity Search feature using Oracle Database 23ai, which converts text into embeddings and uses cosine similarity to find conceptually similar claims. **Why AI Vector Search?** -Keyword searches often miss context. For example, "Affordable Housing Loan" and "Low-Income Home Loan" mean similar things but use different terms. AI Vector Search solves this by converting text into vectors (embeddings) that capture meaning. With Oracle Database 23ai, you will use similarity search to match user queries to loan descriptions based on semantics, not just keywords. If you're ready to dive in, proceed with this lab and start coding. If you prefer a more guided approach with a detailed walkthrough, you can continue to Lab 5b Step-by-step: Code with AI Vector Search for step-by-step instructions. +Unlike keyword search, AI Vector Search captures meaning. For example, '[placeholder]]' and '[placeholder]' use different words but express similar ideas. Oracle Database 23ai converts claim descriptions into vectors, compares their meanings, and ranks results by similarity for more accurate recommendations. Good luck, and enjoy the process. Let’s get coding! @@ -346,9 +346,9 @@ You have successfully: - Connected to Oracle Database 23ai. - Vectorized loan descriptions into embeddings. - Verified the stored data. -- Built a Similarity Search to match user queries to loans. +- Built a Similarity Search to match user queries to treatments. -Fixing the code allowed SeerEquites to recommend loans based on meaning, not keywords, using AI Vector Search in Oracle Database 23ai. You learned to generate embeddings, query vectors, and integrate AI with SQL—skills you can apply to other projects. +Fixing the code allowed Seer Healthcare to recommend treatments based on meaning, not keywords, using AI Vector Search in Oracle Database 23ai. You learned to generate embeddings, query vectors, and integrate AI with SQL—skills you can apply to other projects. ## Learn More diff --git a/dev-ai-app-dev-healthcare/app-architecture/app-architecture.md b/dev-ai-app-dev-healthcare/app-architecture/app-architecture.md index 0ac5630a..aad4003e 100644 --- a/dev-ai-app-dev-healthcare/app-architecture/app-architecture.md +++ b/dev-ai-app-dev-healthcare/app-architecture/app-architecture.md @@ -24,7 +24,7 @@ The SeerEquities loan application runs in an **Oracle Cloud Infrastructure (OCI) - A VM in the public subnet runs two containers, including: - - Open-source Python library for the Loan Approval Demo + - Open-source Python library for the Healthcare Claim Approval Demo - JupterLab as a browser-based development environment @@ -63,7 +63,7 @@ Oracle Database 23ai supports property graphs, which model relationships using v Property graphs simplify working with connected data, such as identifying influencers, predicting trends, or discovering relationships through pattern matching, by offering a more intuitive and efficient way to model and query complex networks. -**Where is it used**: We implemented property graphs in the demo part of this workshop. Loan approval officers can use it to identify potential alternative loans for their customers. +**Where is it used**: We implemented property graphs in the demo part of this workshop. Healthcare Claim Analysts can use it to identify potential alternative treatments for their patients. ## Acknowledgements diff --git a/dev-ai-app-dev-healthcare/build/build.md b/dev-ai-app-dev-healthcare/build/build.md index 70c81da1..59d77e43 100644 --- a/dev-ai-app-dev-healthcare/build/build.md +++ b/dev-ai-app-dev-healthcare/build/build.md @@ -2,7 +2,7 @@ ## Introduction -In this lab, you will build a loan recommendation system powered by Oracle Database 23ai and OCI Generative AI. You will connect to a Oracle Database 23ai, explore customer data, extract relevant insights, and use a Large Language Model (LLM) to generate personalized loan suggestions. The system combines AI with OCI Generative AI to provide personalized loan recommendations. This system integrates AI capabilities with Oracle's robust database technology to deliver targeted financial recommendations. +Here you’ll create an end‑to‑end prior‑authorization system using Oracle Database 23ai and OCI Generative AI. After connecting to clinical datasets, you’ll analyze patient records and leverage an LLM to recommend approval, pend, or denial - complete with evidence citations. By applying Python skills from prior labs, you’ll assemble a fully integrated, AI‑driven healthcare review portal. This lab uses some of the basic coding samples you created in lab 3, such as cursor.execute and more. @@ -15,10 +15,10 @@ To get things started we invite you to watch this video and see the lab in actio ### Objectives -* Build the complete loan approval application as seen in lab 1 -* Use OCI Generative AI to generate personalized loan recommendations +* Build the complete healthcare claim approval application as seen in lab 1 +* Use OCI Generative AI to generate personalized claim recommendations * Use Python to connect to an Oracle Database 23ai instance and run queries -* Explore customer data and extract relevant information +* Explore patient data and extract relevant information ### Prerequisites diff --git a/dev-ai-app-dev-healthcare/cloud-login/cloud-login.md b/dev-ai-app-dev-healthcare/cloud-login/cloud-login.md index 93e74a7b..2601085f 100644 --- a/dev-ai-app-dev-healthcare/cloud-login/cloud-login.md +++ b/dev-ai-app-dev-healthcare/cloud-login/cloud-login.md @@ -2,7 +2,7 @@ ## Introduction -In this lab, we will show you where you can find the login information and how to log in to the **SeerEquities Loan Management application**. +In this lab, we will show you where you can find the login information and how to log in to the **SeerHoldings Healthcare Management application**. Estimated Time: 5 minutes @@ -27,11 +27,11 @@ b. **Time Remaining:** This shows you the remaining time before your access to t ![Click the Start Demo Link](./images/start-demo.png " ") -2. Enter in a username, select the **Loan Officer** role, and click **Login**. +2. Enter in a username, select the **Claims Officer** role, and click **Login**. ![Login](./images/login.png " ") -3. Welcome to the SeerEquities Loan Management application! Congratulations, you are now connected to the demo environment. You can now execute the different tasks and steps for the LiveLabs workshop. +3. Welcome to the SeerHoldings Healthcare Management application! Congratulations, you are now connected to the demo environment. You can now execute the different tasks and steps for the LiveLabs workshop. ![Homepage](./images/app-home.png " ") diff --git a/dev-ai-app-dev-healthcare/codingbasics/codingbasics.md b/dev-ai-app-dev-healthcare/codingbasics/codingbasics.md index 32f4d923..87bb055e 100644 --- a/dev-ai-app-dev-healthcare/codingbasics/codingbasics.md +++ b/dev-ai-app-dev-healthcare/codingbasics/codingbasics.md @@ -2,7 +2,7 @@ ## Introduction -In this hands-on lab, you will learn the basics of working with Oracle Database 23ai using Python. You will connect to the database with both `oracledb` and `pymongo`, create and query tables, and build JSON Duality Views. You will also explore Oracle’s Mongo API to query and update data using a familiar document model. For organizations already using MongoDB-like applications, the Oracle MongoDB API enables a smoother transition and integration with existing code, tools, or drivers—while consolidating everything in Oracle’s converged platform. +In this hands-on lab, you'll use Python to interact with Oracle Database 23ai - connecting with both `oracledb` and `pymongo`, creating tables, querying data, and building JSON Duality Views. You'll also explore Oracle’s Mongo API to query and update data using a document-based approach, making it easy to integrate with existing MongoDB-style apps. This lab is ideal for those with basic Python and Oracle knowledge and builds on the setup from Lab 2. Estimated Time: 30 minutes @@ -187,7 +187,6 @@ Now, that we have established a connection, we can start creating our tables and ![query customers](./images/task3.png " ") -### **Task Summary** Congratulations! You successfully created two new tables with sample data using Python and Oracle Database. @@ -197,7 +196,9 @@ You also created a function that allows you to query your new table which we wil Next, we want to explore how we can use a **JSON Duality View** to query our new table. A JSON Duality View allows us to interact with data as JSON objects, i.e., data is stored as documents. Unlike a regular view, we can also update data in a JSON Duality View. Any updates will be reflected in our original relational tables. We will create a JSON Duality View using our newly created tables `customers` and `orders`. -1. Before we create the **JSON Duality View**, we need to add some **constraints** to our new tables. Copy & paste the following code into a **new cell** and run it. +1. Before we create the **JSON Duality View**, we need to add some **constraints** to our new tables. Constraints define specific rules and conditions that the data in a table must follow. Their main function is to maintain data integrity and ensure the accuracy and reliability of the stored information. + + Copy & paste the following code into a **new cell** and run it. ```python @@ -478,9 +479,9 @@ The final step in our basic coding tour with Python and the Oracle Database 23ai 🔴 **`df.head()`** - This is a method that returns that returns the result of the query including the column names. -## Summary +## Conclusion -In this lab, we learned how to use Python and Oracle's Python driver `oracledb` to interact with Oracle Database 23ai's new features. You learned how to user the `cursor` object to execute SQL queries. Using the `cursor` object, you created a **JSON Duality View** and you even used some JSON functions to query documents using SQL syntax. Then, you also learned how to connect to the database using `pymongo` and retrieve data from a table in the database using **MongoDB syntax**. You created functions to update the **JSON Duality View** and you learned how these updates are also reflected in the underlying relational database tables. +In this lab, we learned how to use Python and Oracle's Python driver `oracledb` to interact with Oracle Database 23ai's new features. You learned how to use the `cursor` object to execute SQL queries. Using the `cursor` object, you created a **JSON Duality View** and you even used some JSON functions to query documents using SQL syntax. Then, you also learned how to connect to the database using `pymongo` and retrieve data from a table in the database using **MongoDB syntax**. You created functions to update the **JSON Duality View** and you learned how these updates are also reflected in the underlying relational database tables. In the next labs, you will see several of the coding principles learned and even more. diff --git a/dev-ai-app-dev-healthcare/connect-to-env/connect-to-env.md b/dev-ai-app-dev-healthcare/connect-to-env/connect-to-env.md index 3a0676a4..b04b6dd5 100644 --- a/dev-ai-app-dev-healthcare/connect-to-env/connect-to-env.md +++ b/dev-ai-app-dev-healthcare/connect-to-env/connect-to-env.md @@ -2,7 +2,7 @@ ## Introduction -In this lab, we will show you how to connect to the development environment. We are using JupyterLab as our development environment, which is a web-based application that allows you to write and run code on your browser. You can use this environment to create notebooks, edit files, and run Python scripts. +In this lab, you'll learn how to access the web-based Jupyter Lab development environment, where you can write and run Python code directly in your browser. You'll log in, create notebooks, and get ready to start coding in a hands-on environment. Estimated Time: 5 minutes @@ -28,9 +28,9 @@ To get things started we invite you to watch this video and see the lab in actio 1. Review the different elements in JupyterLab: - **File browser:** The file browser organizes and manages files within the JupyterLab workspace. It supports drag-and-drop file uploads, file creation, renaming, and deletion. Users can open notebooks, terminals, and text editors directly from the browser. Navigation is fast and intuitive, with breadcrumbs and context menus that surface relevant actions. Users can right-click files to access options like duplicate, shutdown kernel, or open with a specific editor. + **File browser (1):** The file browser organizes and manages files within the JupyterLab workspace. It supports drag-and-drop file uploads, file creation, renaming, and deletion. Users can open notebooks, terminals, and text editors directly from the browser. Navigation is fast and intuitive, with breadcrumbs and context menus that surface relevant actions. Users can right-click files to access options like duplicate, shutdown kernel, or open with a specific editor. - **Launcher:** The launcher offers a streamlined entry point for starting new activities. Users can create Jupyter Notebooks for interactive coding with live code execution, visualizations, and rich markdown. The terminal provides direct shell access, enabling command-line operations within the JupyterLab environment. These two tools form the core of most workflows, supporting both interactive analysis and system-level tasks from a single interface. + **Launcher (2 and 3):** The launcher offers a streamlined entry point for starting new activities. Users can create Jupyter Notebooks for interactive coding with live code execution, visualizations, and rich markdown. The terminal provides direct shell access, enabling command-line operations within the JupyterLab environment. These two tools form the core of most workflows, supporting both interactive analysis and system-level tasks from a single interface. ![JupyterLab Launcher](./images/jupyter.png " ") @@ -48,7 +48,9 @@ You will use a Jupyter Notebook in JupyterLab to build and test the loan recomme ![JupyterLab blocks](./images/block.png " ") +## Conclusion +In this lab you logged into the **IDE Development Environment** for Jupyter Labs. You reviewed different elements from the environment like **File Browser** and **Launcher**. Lastly, you were introduced to concepts to allow you to get started running code smoothly like **executing code blocks** and **outputs and warnings**. ## Acknowledgements * **Author** - Kamryn Vinson diff --git a/dev-ai-app-dev-healthcare/introduction/introduction.md b/dev-ai-app-dev-healthcare/introduction/introduction.md index aba0cb87..240a03da 100644 --- a/dev-ai-app-dev-healthcare/introduction/introduction.md +++ b/dev-ai-app-dev-healthcare/introduction/introduction.md @@ -6,7 +6,7 @@ Generative AI and Oracle Database 23ai makes it possible to unlock deeper insights, faster—directly within your database. -In this hands-on workshop, you’ll build a finance-focused GenAI solution using Oracle Database 23ai. +This hands-on lab introduces you to Generative AI and Vector Search in Oracle Database 23ai, allowing you to build real applications using multiple types of data - no setup required. You'll connect to the database using Python and the Mongo API, run queries, create JSON Duality Views, and implement AI Vector Search and retrieval-augmented generation (RAG). Explore guided healthcare use cases or experiment freely in a browser-based development environment designed to bring Oracle 23ai’s capabilities to life in a healthcare context. To get things started we invite you to watch the introduction video to see the demo in action: @@ -33,11 +33,17 @@ By the end of this workshop, you’ll understand how Oracle Database 23ai helps **Your next breakthrough is just a lab away!** -#### ➡️ **Start with the demo!** +#### ➡️ **Start with the demo! (Lab 1)** -Check out the demo to see how Generative AI and Vector Search in the database, can operate against multiple types of data—empowering you to build real applications and make faster, smarter decisions. The best part? **No setup required**—just start with Lab 1 and follow the guide. You can explore guided use cases or experiment on your own using the demo application. +Check out the demo to see how Generative AI and Vector Search in the database, can operate against multiple types of data—empowering you to build real applications and make faster, smarter decisions. The best part? **No setup required**—just start with Lab 1 and follow the guide. You can explore guided use cases or experiment on your own using the demo application: -#### ➡️ **Start coding!** + * ✅ Login as a Healthcare Claims Analyst + + * ✅ Review patients that have been recommended by the system to be approved or denied for a claim + + * ✅ Upload information to a patient's profile + +#### ➡️ **Start coding! (Labs 2-4)** Lab 2 describes how to get access to our browser-based development environment. Make sure to check it out before continuing to Lab 3 and 4, where you will get hands-on and start interacting with Python with Oracle Database 23ai: @@ -53,15 +59,21 @@ Lab 2 describes how to get access to our browser-based development environment. * ✅ Implement AI Vector Search -In lab 4, you will learn how to **implement a working example of retrieval-augmented generation (RAG) using Oracle Database 23ai.** + * ✅ Implement retrieval-augmented generation (RAG) using Oracle Database 23ai + +#### ➡️ **Start experimenting! (Lab 5)** + +You can also experiment with the code in your own environment. The workshop environment is flexible and allows you try out different things: Start building your own version of the prior authorization application! + + * ✅ Vectorize patient data from one table into another -#### ➡️ **Start experimenting!** + * ✅ Verify the data was vectorized -You can also experiment with the code in your own environment. The workshop environment is flexible and allows you try out different things: Start building your own version of the loan approval application! + * ✅ Implement a similarity search ### Objectives -* Learn how Oracle Database 23ai capabilities can be used to automate and enhance the loan approval process +* Learn how Oracle Database 23ai capabilities can be used to automate and enhance the prior authorization process, justifying medical-necessity decisions * Start coding with Python and Oracle's Python driver `oracledb` diff --git a/dev-ai-app-dev-healthcare/user-story/user-story.md b/dev-ai-app-dev-healthcare/user-story/user-story.md index c70267d1..ce88e2e5 100644 --- a/dev-ai-app-dev-healthcare/user-story/user-story.md +++ b/dev-ai-app-dev-healthcare/user-story/user-story.md @@ -2,7 +2,7 @@ ## Introduction -In this lab, you will step into the role of a loan officer using a next-generation loan approval application powered by Oracle Database 23ai. You will work with real loan applications and see how Generative AI, Vector Search, and Graph analytics replace manual review with faster, AI-driven decision-making. +Step into the role of a provider or claims analyst authorizing an MRI and specialty medication using an AI-powered application. Built on Oracle Database 23ai, this demo highlights how Generative AI, AI Vector Search, Graph Analytics, and JSON Duality Views come together to streamline documentation, surface policy-aligned insights, and produce regulator-compliant outputs—all without data movement **Disclaimer**: Please note that your results may vary. The information provided is generated by OCI Generative AI services, and your outcomes may differ from those presented. @@ -12,7 +12,7 @@ Estimated Lab Time: 30 minutes In this lab, you will: -* Review how the Seer Equities Loan approval app incorporates the use of JSON Duality Views, Graph analytics, and other converged database features, all without requiring complex data movement or separate systems. +* Review how the Seer Healthcare prior authorization app incorporates the use of JSON Duality Views, Graph analytics, and other converged database features, all without requiring complex data movement or separate systems. ### Prerequisites @@ -30,56 +30,56 @@ This lab assumes you have: ![Login](./images/login.png =50%x*) -3. Welcome to the SeerEquities Loan Management application! Congratulations, you are now connected to the demo environment. You can now execute the different tasks for this Lab. +3. Welcome to the Seer Healthcare Prior Authorization application! Congratulations, you are now connected to the demo environment. You can now execute the different tasks for this Lab. ![Homepage](./images/app-home.png =50%x*) -## Task 2: Demo - Customer with strong credit score +## Task 2: Demo - Patient with low authorization risk score -In this first example, you will use the application to approve a customer with strong credit score. The first user on your to-do list is James Smith. +In this first example, you will use the application to approve a patient with a low authorization risk score. The first user on your to-do list is James Smith. -1. On the Dashboard page, from the pending review list, select the Customer ID for **James Smith**. +1. On the Dashboard page, from the pending review list, select the Patient ID for **James Smith**. ![Select James Smith](./images/james-smith.png =50%x*) -2. Opening James Smith’s profile reveals his loan application details—name, location, requested amount, debt, and credit score. +2. Opening James Smith’s profile reveals his prior authorization request details—name, physcian, physician's notes, and patient condition. ![James Smith AI generated recommendations](./images/james-smith-ai.png =50%x*) -3. At the bottom of James Smith’s profile, you will find the **AI Loan Guru**—a chatbot built on Oracle Database 23ai and Vector search. When prompted, the system uses **RAG** to generate a response. It converts the question and loan data into embeddings, performs a similarity search, and then uses the **GenAI service** to turn the enriched context into a clear, natural language answer. If the customer calls with a question, you can quickly enter it into the AI Loan Guru to generate a relevant response. +3. At the bottom of James Smith’s profile, you will find the **AI Healthcare Guru**—a chatbot built on Oracle Database 23ai and Vector search. When prompted, the system uses **RAG** to generate a response. It converts the question and healthcare data into embeddings, performs a similarity search, and then uses the **GenAI service** to turn the enriched context into a clear, natural language answer. If the customer calls with a question, you can quickly enter it into the AI Healthcare Guru to generate a relevant response. - **Copy** the question below into the AI chatbot and press **Enter**. What does the AI Loan Guru recommend? + **Copy** the question below into the AI chatbot and press **Enter**. What does the AI Healthcare Guru respond? ```text - What would be the top 0% down payment loans? + Why was this request denied? ``` ![James Smith chatbot](./images/james-smith-chatbot.png =50%x*) ->💡 In Oracle Database 23ai, **AI Vector Search** allows you to combine your business data with the Large Language Model (LLM) to reduce hallucinations and get accurate answers from your data. +>💡 In Oracle Database 23ai, **AI Vector Search** allows you to combine your business data with a Large Language Model (LLM) to reduce hallucinations and get accurate answers from your data. 4. Select the **Navigate To Decisions** button. ![James Smith Decision](./images/james-smith-decision.png =50%x*) - After navigating to the decisions page, the AI evaluation runs in the background. It analyzes James’s profile and matches it against available loan options in the database. A custom AI prompt ensures the system uses only internal data—never the internet. In this case, the AI returns three loan options, each with a clear explanation. All options are displayed alongside the AI’s final recommendation: approval. + After navigating to the decisions page, the AI evaluation runs in the background. Clinical documents (e.g., EHR notes, radiology reports, lab results) are stored in Oracle Database 23ai and modeled with JSON Duality Views. These documents are vectorized and searched using AI Vector Search to identify semantically relevant prior cases and policy criteria. The results are passed to OCI Generative AI, which uses retrieval-augmented generation (RAG) to generate a natural-language explanation, formatted for both internal review and external (regulatory or provider) communication. -5. In the **Select Final Approved Loan Option** section, the available loan options are displayed. The loan officer can choose to request more information to refine the offer, but for this scenario, we will proceed by selecting one of the suggested loans. +5. In the **Select Final Prior Authorization Option** section, the available options are displayed. If more data is needed, the reviewer can upload documents, add missing codes, and trigger a re-evaluation. The system dynamically updates the risk graph, adjusts the recommendation, and generates a decision letter sent directly to the provider ![James Smith Select Final Approved Loan](./images/james-smith-select-final-loan.png =50%x*) -6. Select the AI-recommended loan labeled as the military veteran loan with the shortest time to close. In this example, that is Loan 26—the only military veteran option shown. +6. Select the AI-recommended authorization decision. In this example, that is **Approved**. >Please note that your results may vary. The information provided is generated by OCI Generative AI services, and your outcomes may differ from those presented. ![James Smith Select Loan 3](./images/james-smith-select-loan.png =50%x*) -7. Set the final loan status to **Approved**, then click **Save Final Approval & Loan Status** to complete the process. +7. Set the final authorization status to **Approved**, then click **Save Final Approval & Authorization Status** to complete the process. - The loan status has been updated to 'Approved' and saved to the customer profile. + The authorization status has been updated to 'Approved' and saved to the patient profile. ![James Smith Save Loan Status](./images/james-smith-save.png =50%x*) @@ -91,7 +91,7 @@ In this first example, you will use the application to approve a customer with s ![James Smith Download PDF](./images/james-smith-download-pdf.png =50%x*) -10. Display the message the customer would see by opening the downloaded PDF. +10. Display the message the patient would see by opening the downloaded PDF. ![Open James Smith PDF](./images/open-james-smith-pdf.png =50%x*) @@ -99,38 +99,38 @@ In this first example, you will use the application to approve a customer with s ![Return to Dashboard](./images/james-return-dashboard.png =50%x*) -12. Expand **View Approved Customers**. We can see that James Smith has been removed from the Pending Review Loan Requests list and has been added to the Approved Loan Requests list. +12. Expand **View Approved Patients**. We can see that James Smith has been removed from the Pending Prior Authorization Requests list and has been added to the Approved Authorization Requests list. ![James Smith Approved List](./images/james-smith-approved-list.png =50%x*) **Conclusion** -Once you select and save one of the 3 loans recommended by the AI: +Once you select and save one of the 3 authorization options recommended by the AI: -✅ The customer's loan status is updated. +✅ The patient's authorization status is updated. ✅ A finalized PDF decision document is generated. ✅ The dashboard reflects the change in real-time — marking James as Approved. -Congratulations, you have just approved your first customer loan! Proceed to the next task. +Congratulations, you have just approved your first patient authorization! Proceed to the next task. -## Task 3: Demo - Denying a customer with high risk +## Task 3: Demo - Denying a patient with high risk In this example, you will navigate the application to review a customer and deny them as part of the exercise. The second user on your to-do list is James Woods. -1. On the Dashboard page, from the pending review list, select the Customer ID for **James Woods**. +1. On the Dashboard page, from the pending review list, select the Patient ID for **James Woods**. ![Select James Woods](./images/james-woods.png =50%x*) -2. Opening James Woods’s profile displays his loan application details. Within a few seconds, the AI generates automatically recommendations. In this case, the system evaluates a less favorable profile and identifies key risk factors. +2. Opening James Woods’s profile displays his patient history details. Within a few seconds, the AI generates automatically recommendations. In this case, the system evaluates a less favorable profile and identifies key risk factors. - This customer has: + This customer has: [Placeholder] - * **$5000 in income** - * A **low credit score** - * And is requesting a **high loan amount** + * **several case similarities** + * A **minimal policy** + * And is requesting a **high lab threshold** - Despite the risk factors, the AI evaluates the profile and suggests next steps. In this case, it recommends a denial—but also provides clear, actionable guidance to help the customer improve their chances of approval in the future. + The AI evaluates the profile and suggests next steps. In this case, it recommends a denial—but also provides clear, actionable guidance to help the patient improve their chances of approval in the future. ![James Woods AI generated recommendations](./images/james-woods-ai.png =50%x*) @@ -141,23 +141,23 @@ In this example, you will navigate the application to review a customer and deny >⁉️ **What are two reasons that the AI decided to deny this applicant?** ⁉️ -4. Expand **View Dot Plot: Client-Loan Relationships** to view the graph. +4. Expand **View Dot Plot: Diagnosis-Policy Relationships** to view the graph. ![James Woods Graph](./images/expand-graph.png =50%x*) - On the decision page, the loan officer can use **Operational Property Graph** to explore near-approval loan scenarios. Built with **Oracle Graph**, this feature visually maps loans just out of reach and highlights what adjustments—like increasing income or reducing debt—could improve eligibility. + On the decision page, the healthcare claim reviewer can use **Operational Property Graph** to explore near-approval authorization scenarios. Built with **Oracle Graph**, this feature visually maps authorizations just out of reach and highlights what adjustments—like adding missing codes or re-evaluation could improve eligibility. ![James Woods Graph](./images/james-woods-graph2.png =50%x*) ->💡 In Oracle Database 23ai, **Property Graph** allows you to treat your data like a network of connected points, where each point (called a node) and each link (called an edge) has its own details or properties. This setup helps you run graph analytics, like finding important connections or patterns, directly within the database. +>💡 In Oracle Database 23ai, **Property Graph** allows you to treat your data like a network of connected points, where each point (called a node) and each link (called an edge) has its own details or properties. This setup helps you run graph analytics, like to find important connections or patterns, directly within the database. -5. The loan status is set to **Denied**. Click the **Save Final Approval & Loan Status** button. +5. The authorization status is set to **Denied**. Click the **Save Final Approval & Authorization Status** button. - The loan status has been updated to 'Denied' and saved to the customer profile. + The authorization status has been updated to 'Denied' and saved to the patient profile. ![James Woods Save Loan Status](./images/james-woods-save.png =50%x*) -6. Press the **Generate Decision as PDF** button to save the AI responses and proceed to the final loan disposition. +6. Press the **Generate Decision as PDF** button to save the AI responses and proceed to the final authorization disposition. ![James Woods PDF](./images/james-woods-pdf.png =50%x*) @@ -165,7 +165,7 @@ In this example, you will navigate the application to review a customer and deny ![James Woods Download Recommendation](./images/james-woods-download.png =50%x*) -8. Display the message the customer would see by opening the downloaded PDF. +8. Display the message the patient would see by opening the downloaded PDF. ![Open James Woods PDF](./images/open-james-woods-pdf.png =50%x*) @@ -173,23 +173,23 @@ In this example, you will navigate the application to review a customer and deny ![Return to Dashboard](./images/woods-return-dashboard.png =50%x*) -10. Expand **View Denied Customers**. You will see that James Woods has been moved from the **Pending Review Loan Requests** list to the **Denied Loan Requests** list. +10. Expand **View Denied Patients**. You will see that James Woods has been moved from the **Pending Prior Authorization Requests** list to the **Denied Authorization Requests** list. ![James Woods Denied List](./images/james-woods-denied-list.png =50%x*) **Conclusion** -Congratulations, you have finished reviewing a customer with high financial risk! Proceed to the next task. +Congratulations, you have finished reviewing a patient with high fraudulent risk! Proceed to the next task. -## Task 4: Demo - Update customer details +## Task 4: Demo - Update patient details -Lastly, let’s explore how the system uses JSON Duality Views to handle profile updates. In this task, you will edit a customer’s details. In this example, the customer was asked to submit updated proof of income. +Lastly, let’s explore how the system uses JSON Duality Views to handle profile updates. In this task, you will edit a patient's details. In this example, the patient was asked to submit updated lab results. -1. On the Dashboard page, from the **Pending Review Loan Requests** list, select the customer ID for **Alex Anderson**. +1. On the Dashboard page, from the **Pending Prior Authorization Requests** list, select the patient ID for **Alex Anderson**. ![Select Alex Anderson](./images/select-alex-anderson.png =50%x*) -2. We will upload a document to update Alex's income. Before uploading the document, note that the customer’s income is currently listed as $25,000. On the Customer Details page, click the **Upload Document** button. +2. We will upload a document to update Alex's lab results. Before uploading the document, note that the patient's labs is currently listed as **none**. On the Patient Details page, click the **Upload Document** button. ![Upload Document](./images/upload-document.png =50%x*) @@ -199,11 +199,11 @@ Lastly, let’s explore how the system uses JSON Duality Views to handle profile >💡 **JSON Duality Views** in 23ai let's you update unstructured data in an easy, high-level format while automatically handling the technical details behind the scenes. This makes it faster and simpler to work with messy data and connect it to structured systems. -4. The customer profile has been updated. +4. The patient profile has been updated. ![Customer Profile Updated](./images/profile-updated.png) -5. Refresh the page and note that the income has been updated to $65,000. Thanks to JSON Transform and JSON Duality Views, only the relevant field is modified — leaving the rest of the profile UNTOUCHED. +5. Refresh the page and note that the labs has been updated to [placeholder]. Thanks to JSON Transform and JSON Duality Views, only the relevant field is modified — leaving the rest of the profile UNTOUCHED. ![Customer Profile Updated](./images/income-updated.png) @@ -211,25 +211,25 @@ Lastly, let’s explore how the system uses JSON Duality Views to handle profile Once the document is uploaded: -✅ The system automatically detects the new income data. +✅ The system automatically detects the new lab data. -✅ Then their profile will be updated from $25,000 to $65,000. +✅ Then their profile will be updated from **none** to [placeholder]. ✅ And thanks to JSON Transform and JSON Duality Views, only the relevant field is modified — leaving the rest of the profile UNTOUCHED. ## Summary -In conclusion our Loan Management App was able to leverage Oracle database 23ai technologies such as **AI Vector Search, Property Graph and JSON Duality Views** to: +In conclusion our Prior Authorizations Management App was able to leverage Oracle database 23ai technologies such as **AI Vector Search, Property Graph and JSON Duality Views** to: ✅ Automate profile evaluations -✅ Provide AI-driven loan recommendations by using an RAG model powered by a Oracle Database 23ai's AI Vector Search and OCI Generative AI service +✅ Provide AI-driven authorization recommendations by using an RAG model powered by a Oracle Database 23ai's AI Vector Search and OCI Generative AI service ✅ Enable seamless profile updates with JSON Duality Views -✅ And empower loan officers with actionable insights through Operational Property Graphs +✅ And empower healthcare claim reviewers with actionable insights through Operational Property Graphs -By combining these advanced tools, the application enables faster, smarter decisions and delivers clear guidance on how customers can improve their eligibility. +By combining these advanced tools, the application enables faster, smarter decisions and delivers clear guidance on how patients can improve their eligibility. **Next:** How about learning how the application was implemented in Python? Continue with the next labs and start developing! diff --git a/dev-ai-app-dev-healthcare/workshops/sandbox/manifest.json b/dev-ai-app-dev-healthcare/workshops/sandbox/manifest.json index bf458e31..52bbe611 100644 --- a/dev-ai-app-dev-healthcare/workshops/sandbox/manifest.json +++ b/dev-ai-app-dev-healthcare/workshops/sandbox/manifest.json @@ -1,5 +1,5 @@ { - "workshoptitle": "Build a GenAI-Powered Financial Services Loan Approval Application with Oracle Database 23ai", + "workshoptitle": "Build a GenAI App on Oracle Database 23ai – Healthcare Edition", "help": "livelabs-help-database_us@oracle.com", "tutorials": [ { diff --git a/dev-ai-app-dev-healthcare/workshops/tenancy/manifest.json b/dev-ai-app-dev-healthcare/workshops/tenancy/manifest.json index 104133f3..1c09880b 100644 --- a/dev-ai-app-dev-healthcare/workshops/tenancy/manifest.json +++ b/dev-ai-app-dev-healthcare/workshops/tenancy/manifest.json @@ -1,5 +1,5 @@ { - "workshoptitle": "Build a GenAI-Powered Financial Services Loan Approval Application with Oracle Database 23ai", + "workshoptitle": "Build a GenAI App on Oracle Database 23ai – Healthcare Edition", "help": "livelabs-help-database_us@oracle.com", "tutorials": [ { diff --git a/dev-ai-app-dev-retail/workshops/sandbox/manifest.json b/dev-ai-app-dev-retail/workshops/sandbox/manifest.json index bf458e31..f6544796 100644 --- a/dev-ai-app-dev-retail/workshops/sandbox/manifest.json +++ b/dev-ai-app-dev-retail/workshops/sandbox/manifest.json @@ -1,5 +1,5 @@ { - "workshoptitle": "Build a GenAI-Powered Financial Services Loan Approval Application with Oracle Database 23ai", + "workshoptitle": "Build a GenAI App on Oracle Database 23ai – Retail Edition", "help": "livelabs-help-database_us@oracle.com", "tutorials": [ { diff --git a/dev-ai-app-dev-retail/workshops/tenancy/manifest.json b/dev-ai-app-dev-retail/workshops/tenancy/manifest.json index 104133f3..ef428933 100644 --- a/dev-ai-app-dev-retail/workshops/tenancy/manifest.json +++ b/dev-ai-app-dev-retail/workshops/tenancy/manifest.json @@ -1,5 +1,5 @@ { - "workshoptitle": "Build a GenAI-Powered Financial Services Loan Approval Application with Oracle Database 23ai", + "workshoptitle": "Build a GenAI App on Oracle Database 23ai – Retail Edition", "help": "livelabs-help-database_us@oracle.com", "tutorials": [ {