COURSE OUTLINE

Day 1: AI-Powered Development & Advanced Prompting

Focus: Transforming the developer workflow.  

• The 2026 Developer Stack: Navigating GPT-5.5, Gemini 3.0, and specialized coding models (Gemma-Code).
• Advanced Prompt Patterns for Engineers:
• Persona Pattern: Assigning deep domain expertise (e.g., "Act as a Senior SRE").
• Chain-of-Thought (CoT): Forcing models to reason through complex logic before coding.
• AI-Driven Refactoring: Modernizing legacy codebases and automated technical debt identification.
• Hands-on: Refactoring a monolithic service into a clean, documented microservice architecture using ChatGPT.

Day 2: Architecting with LLMs & APIs

Focus: Building software that "thinks" as a feature.

• The AI API Layer: Deep dive into the OpenAI and Vertex AI SDKs.  
• Structured Outputs & Tool Use: Ensuring the model returns valid JSON and knows when to call an external function.
• State & Memory: Managing conversation history and long-term memory in distributed systems.
• Hands-on: Building a CLI tool that interacts with your local filesystem to automatically fix linter errors and commit changes.

Day 3: Context Engineering & Advanced RAG

Focus: Teaching models about your private codebase.

• The Context Window Challenge: Managing 1M+ token contexts and the "Lost in the Middle" problem.
• Building the RAG Pipeline:
• Embeddings & Vectors: Using Vertex AI Vector Search to index entire repositories.
• Hybrid Retrieval: Combining semantic search with keyword search for technical precision.
• Hands-on Lab: Creating a "Codebase Expert" that can answer architectural questions across 50+ microservices.

Day 4: Agentic Workflows & Orchestration

Focus: Moving from chatbots to autonomous software agents.

• Agent Design Patterns: ReAct (Reason + Act), Plan-and-Solve, and Multi-Agent collaboration.
• Orchestration Frameworks: Deep dive into LangGraph and Google’s Agent Development Kit (ADK).
• Error Handling in AI: Designing fallbacks for when an agent enters a "hallucination loop."
• Hands-on Lab: Building an "Auto-QA Agent" that writes, runs, and fixes its own unit tests until they pass.

Day 5: LLMOps, Security, and Production

Focus: Shipping reliable, enterprise-grade AI.

• LLMOps Fundamentals: Model versioning, prompt evaluation (LLM-as-a-judge), and A/B testing.
• Security & Red Teaming: Defending against Indirect Prompt Injection and data exfiltration.
• Cost & Latency Optimization: Implementing prompt caching and choosing between Pro vs. Flash models.
• Capstone Presentation: Deploying a "Self-Healing App" that monitors its own logs and suggests fixes via Gemini.