Building Generative AI Applications

Objective: The student is able to build AI applications to solve problems, beyond manual copy-pasting into commonly available chat interfaces like ChatGPT and Gemini.

8-Day AI Engineering Course Outline

Day 1: Foundations

• Prompting (day_1a_prompting.ipynb)
  • Zero-Shot Prompting
  • Few-Shot Prompting
  • Chain-of-Thought Prompting
  • System Instructions
  • Code Generation and Execution
• Evaluation and Structured Output (day_1b_evaluation-and-structured-output.ipynb)
  • Structured Output
  • Pointwise Evaluation
  • Pairwise Evaluation
  • Evaluation Methods

Day 2: Embeddings & RAG

• Embeddings and Similarity Scores (day_2_similarity.ipynb)
  • Understanding Embeddings
  • Semantic Similarity Calculations
• Classifying Embeddings with PyTorch (day_2_classifier_pytorch.ipynb)
  • Classification tasks using embeddings
  • Building classification models with PyTorch
• Document Q&A with RAG (day_2_qa_rag.ipynb)
  • Retrieval Augmented Generation implementation
  • Using ChromaDB for vector storage

Day 3: Function Calling & Grounding

• Function Calling with the Gemini API (day_3_function_calling.ipynb)
  • Function Calling basics
  • Automatic Function Calling
  • Database integration with function calling
• Google Search Grounding (day_3_search_grounding.ipynb)
  • Grounding with Google Search
  • Inspecting grounding metadata
  • Adding inline citations

Day 4: From Prompt to Action & Agent Architectures

• From Prompt to Action (day_4a_from_prompt_to_action.ipynb)
  • Building your first AI agent with ADK
  • Agent Development Kit (ADK) setup
  • Creating agents with tools
• Multi-Agent Systems & Workflow Patterns (day_4b_agent_architectures.ipynb)
  • Sequential Agents
  • Parallel Agents
  • Loop Agents
  • Multi-agent orchestration patterns

Day 5: Agent Tools & Best Practices

• Agent Tools (day_5a_agent_tools.ipynb)
  • Custom Function Tools
  • Agent Tools (using agents as tools)
  • Built-in Code Executor
  • Tool types overview
• Agent Tool Patterns and Best Practices (day_5b_agent_tools_best_practices.ipynb)
  • Model Context Protocol (MCP) integration
  • Long-Running Operations (LRO)
  • Human-in-the-loop approvals
  • Resumable workflows

Day 6: Sessions & Memory

• Memory Management - Part 1: Sessions (day_6a_agent_sessions.ipynb)
  • Session Management
  • Persistent Sessions with DatabaseSessionService
  • Context Compaction
  • Session State management
• Memory Management - Part 2: Memory (day_6b_agent_memory.ipynb)
  • Long-term memory with MemoryService
  • Memory ingestion and retrieval
  • Automatic memory storage with callbacks
  • Memory consolidation concepts

Day 7: Observability & Evaluation

• Agent Observability (day_7a_agent_observability.ipynb)
  • Logs, Traces & Metrics
  • Debugging with ADK Web UI
  • LoggingPlugin for production
  • Custom plugins and callbacks
• Agent Evaluation (day_7b_agent_evaluation.ipynb)
  • Interactive evaluation with ADK Web UI
  • Systematic evaluation with test cases
  • Tool trajectory and response metrics
  • Regression testing

Day 8: Agent2Agent Communication & Deployment

• Agent2Agent (A2A) Communication (day_8a_agent2agent_communication.ipynb)
  • A2A Protocol overview
  • Exposing agents via A2A
  • Consuming remote agents
  • Cross-framework and cross-organization integration
• Agent Deployment (day_8b_agent_deployment.ipynb)
  • Deploying to Vertex AI Agent Engine
  • Production-ready agent configuration
  • Testing deployed agents
  • Long-term memory with Vertex AI Memory Bank
  • Cost management and cleanup