AI engineering with Python — Anthropic SDK, RAG, evals, agents
ChatGPT API tutorials end at 'hello world'. We start at the boring middle — versioning, evals, retrieval — where shipped AI features actually live.
Most 'AI engineering' content stops at calling the API. The hard part — and the part that gets you hired — is everything after the first response: chunking strategies for retrieval, eval datasets that catch regressions, prompt versioning that survives a team of three, agent loops that don't burn $400 in token spend before noon. That's where the AI Engineering track lives.
100 lessons paired with a 100-day path: Anthropic SDK setup, adaptive thinking, tool use, RAG end-to-end (chunking → embeddings → BM25 + dense → reranking), structured outputs, planner-worker-synthesizer agent patterns, evals on real datasets, fine-tuning trade-offs. By the end you have two portfolio projects — a production RAG system and a tool-using assistant — both deployable.