The GenAI Agents Course: Building AI Agents from Simple Scripts to Production Systems

Topics covered

• •Understand what turns an LLM call into an "agent": the observe → reason → act loop, state, and tools.
• •Build a conversational agent with session history management.
• •Build a QA agent and a natural-language data analysis agent.

Topics covered

• •Master LangGraph: StateGraph, nodes, edges, state, graph compilation and visualization.
• •Understand the Model Context Protocol (MCP) as an open standard for connecting LLMs to external resources.
• •Connect an agent to the live internet: search via DuckDuckGo and summarize the results.

Topics covered

• •Apply LangGraph to real business scenarios: customer support, text grading, planning, HR, and compliance.
• •Master conditional routing: categorize the request → route to different branches → escalate.
• •Learn to build multi-step conversational applications that collect user input.

Topics covered

• •Build multimodal pipelines: text → image (DALL-E), text → speech (TTS), text → music (MIDI).
• •Master asynchronous processing (asyncio, aiohttp) for parallel content generation.
• •Learn to adapt one piece of content for multiple platforms while preserving the brand voice.

Topics covered

• •Build news pipelines: search → extract → summarize (NewsAPI, Tavily, BeautifulSoup).
• •Master audio processing: transcribing calls with Whisper and NLP analysis.
• •Integrate real-time data (weather APIs) with severity-based routing and human-in-the-loop.

Topics covered

• •Understand when several specialized agents beat a single general-purpose one, and when they don't.
• •Master four multi-agent frameworks: LangGraph, AutoGen, CrewAI, and OpenAI Swarm.
• •Study archetypal roles: coordinator/supervisor, planner, executor, critic/QA.

Topics covered

• •Master advanced RAG: knowledge-graph RAG (LightRAG), NER-enriched retrieval, local LLMs (Ollama).
• •Build an agent with three-tier memory (semantic, episodic, procedural) using LangMem.
• •Implement self-healing: automatic detection, diagnosis, and fixing of code errors, with a bug memory in ChromaDB.

Topics covered

• •Master automated testing of agentic systems: generating test cases, inspecting graphs, finding vulnerabilities.
• •Build E2E tests from natural-language instructions (LangGraph + Playwright).
• •Implement document intake: converting office formats into LLM-ready markdown as a tool call.