Advanced Topics in Analytics II - Generative AI

Instructor: Sergio A. Mora Pardo

• email: sergioa.mora@javeriana.edu.co
• github: sergiomorapardo

This course is the second part of Advanced Topics in Analytics, moving from classic NLP and Graph Learning into the world of Generative AI. Students will learn how to build real applications powered by Large Language Models (LLMs), covering prompt engineering, working with commercial LLM APIs (OpenAI, Anthropic/Claude), embeddings and vector databases, Retrieval-Augmented Generation (RAG) from basics to advanced techniques (re-ranking, hybrid search, query transformation), AI agents (ReAct, tool use, function calling), orchestration frameworks such as LangChain and LangGraph, the Model Context Protocol (MCP), multi-agent systems, evaluation of LLM applications (LLM-as-judge, RAGAS) and LLMOps for deployment. The course is project-oriented, with emphasis placed on writing software implementations that solve real-world problems end to end.

Motivation

The original course (Advanced Topics in Analytics) taught students to understand language models — the NLP, embeddings and transformer foundations of how they work. This course takes the next step: it teaches students to build real systems that use those models, moving them from consumers of AI to engineers of AI.

Why this matters:

• Employability — RAG, LLM evaluation and LLMOps are precisely the skills companies are hiring GenAI and AI-application engineers for today.
• Higher-order skills — students progress from merely applying AI to designing, evaluating and creating AI systems.
• Transferable, vendor-independent engineering — the fundamentals of RAG, evaluation and agent architecture outlast any single model, provider or framework.

And one honest note: students will also learn when not to reach for a heavy framework. Recognizing that a direct API call is enough for the task at hand is itself a mark of good engineering judgment.

Learning Outcomes

By the end of the course, students will be able to design, connect (to real data), evaluate and deploy LLM-powered applications.

Module	Learning outcome
Intro to LLMs & Prompt Engineering	Explain how LLMs work and craft effective, structured prompts.
Working with LLM APIs	Integrate OpenAI and Anthropic APIs; manage tokens, parameters and streaming.
Embeddings & Vector Databases	Represent text as embeddings and store/query them in FAISS and Chroma.
RAG fundamentals	Build a retrieval-augmented pipeline that grounds an LLM on external data.
Advanced RAG	Improve retrieval with re-ranking, hybrid search and query transformation.
AI Agents fundamentals	Design agents that reason and use tools (ReAct, function calling).
LangChain	Compose LLM applications with chains, memory and tools.
LangGraph	Build stateful, multi-step agent workflows with explicit control flow.
MCP	Connect agents to external tools and data through the Model Context Protocol.
Multi-agent systems	Orchestrate multiple collaborating agents for complex tasks.
Evaluation of LLM apps	Measure quality with RAGAS and LLM-as-judge instead of relying on vibe-checks.
LLMOps & Deployment	Deploy, monitor and manage the cost/latency of LLM apps in production.

Requiriments

• Python version >= 3.10;
• LangChain, framework for building LLM applications;
• LangGraph, library for stateful, multi-step agents;
• langchain-openai, OpenAI integration for LangChain;
• langchain-anthropic, Anthropic/Claude integration for LangChain;
• openai, the official OpenAI Python SDK;
• anthropic, the official Anthropic (Claude) Python SDK;
• chromadb, open-source embedding/vector database;
• faiss-cpu, library for efficient similarity search;
• sentence-transformers, embeddings for sentences and paragraphs;
• tiktoken, fast BPE tokenizer for OpenAI models;
• ragas, evaluation framework for RAG pipelines;
• python-dotenv, management of environment variables and API keys;
• Jupyter, with the additional libraries required for the notebook interface.

A good, easy to install option that supports Mac, Windows, and Linux, and that has all of these packages (and much more) is the Anaconda. All requirements are also listed in requirements.txt.

You will need API keys to work with commercial LLMs. Copy .env.example to .env and fill in your OPENAI_API_KEY and ANTHROPIC_API_KEY.

GIT!! Unfortunatelly out of the scope of this class, but please take a look at these tutorials

Evaluation

• 50% Project
• 40% Exercises
• 10% Class participation

Schedule

Sessions are weekly on Saturdays, Aug 1 - Oct 17, 2026.

1. Intro to LLMs & Prompt Engineering

Date	Session	Notebooks/Presentations	Exercises
August 1st, 2026	Introduction to Large Language Models & Prompt Engineering	L1 - Intro to LLMs

2. Working with LLM APIs (OpenAI, Anthropic/Claude)

Date	Session	Notebooks/Presentations	Exercises
August 8th, 2026	Calling LLM APIs: OpenAI and Anthropic/Claude

3. Embeddings & Vector Databases (FAISS, Chroma)

Date	Session	Notebooks/Presentations	Exercises
August 15th, 2026	Embeddings and Vector Databases with FAISS and Chroma

4. RAG basics

Date	Session	Notebooks/Presentations	Exercises
August 22nd, 2026	Retrieval-Augmented Generation fundamentals

5. Advanced RAG (re-ranking, hybrid search, query transformation)

Date	Session	Notebooks/Presentations	Exercises
August 29th, 2026	Advanced RAG: re-ranking, hybrid search and query transformation

6. AI Agents fundamentals (ReAct, tool use, function calling)

Date	Session	Notebooks/Presentations	Exercises
September 5th, 2026	AI Agents: ReAct, tool use and function calling

7. LangChain

Date	Session	Notebooks/Presentations	Exercises
September 12th, 2026	Building LLM applications with LangChain

8. LangGraph (stateful/multi-step agents)

Date	Session	Notebooks/Presentations	Exercises
September 19th, 2026	Stateful, multi-step agents with LangGraph

9. MCP - Model Context Protocol

Date	Session	Notebooks/Presentations	Exercises
September 26th, 2026	Connecting tools and data with the Model Context Protocol

10. Multi-agent systems

Date	Session	Notebooks/Presentations	Exercises
October 3rd, 2026	Designing and orchestrating multi-agent systems

11. Evaluation of LLM apps (LLM-as-judge, RAGAS)

Date	Session	Notebooks/Presentations	Exercises
October 10th, 2026	Evaluating LLM applications: LLM-as-judge and RAGAS

12. LLMOps & Deployment

Date	Session	Notebooks/Presentations	Exercises
October 17th, 2026	LLMOps: deploying and monitoring LLM applications

Interest Links 🔗

Module	Topic	Material
LangChain	Official Documentation	LangChain Docs
LangGraph	Official Documentation	LangGraph Docs
MCP	Model Context Protocol	modelcontextprotocol.io
GenAI	Short Courses	DeepLearning.AI Short Courses

---

📚 This course is a continuation ("part 2") of Advanced Topics in Analytics, which covered MLOps, Deep Learning, NLP and Graph Learning.