About This Course

Advanced Topics in AI covers what building one RAG pipeline, one semantic layer, or one knowledge graph doesn't teach you: fine-tuning versus retrieval, agent design beyond a single tool call, the cost/latency/quality trade-offs production systems actually get pulled between, and the security and governance failures that only show up once an agent can act on its own. You'll work through curated modules from Factored's LLMOps course and DeepLearning.AI, then prove it on five graded quizzes.

Requirements

• Completed the RAG course, or equivalent experience with chunking, embeddings, vector DBs, and baseline RAG evaluation
• Completed Knowledge Graphs & GraphDBs and/or Semantic Layer & Ontology is helpful but not required
• Comfortable writing Python — used throughout the fine-tuning, agent, and evaluation modules
• Docker installed and running, for the hands-on labs linked from each module
• An LLM API key, needed for the agent, evaluation, and guardrails sections

What You'll Learn

• How to monitor, log, and trace an LLM/RAG system once it's live — not just build one that works in a demo
• When to fine-tune a model instead of retrieving better context, and the practical mechanics of doing it
• The cost, latency, and quality trade-offs that pull almost every production AI decision in different directions, including self-hosted inference serving with vLLM as a cheaper alternative to managed APIs
• Agent design patterns beyond a single tool call — coordination, memory, the A2A protocol — and how to secure an agent that can act on its own: guardrails, governance, and the OWASP Top 10 for agentic applications

Hands-On

Each module points to real, optional labs and projects from the linked courses — deploying RAG monitoring in AWS, running an LLM fine-tuning and evaluation project, building a multi-agent stock market advisor. The course itself is checked with five graded quizzes, each gated on actually completing the paired external coursework first, not just skimming it.