🤖 Python AI engineering
Python AI engineering — build with Claude, ship like the US dev community does.
105 interactive lessons: Claude + LLM APIs, tool use + function calling, RAG, agent loops, evals, prompt caching, voice agents, multi-agent orchestration. Production AI patterns from day 1.
7-day free trial · Cancel in 1 click · $89/year for everything
What you'll cover
API fundamentals
Anthropic Messages API + OpenAI Chat Completions, async clients, streaming with backpressure, token counting, cost math per call. The non-glamorous foundations everyone skips.
Tool use + function calling
Defining tools as JSON schemas, the iter-loop with safety caps, parallel tool calls, fallback when the model picks the wrong tool. Includes a customer-support agent capstone.
RAG
Embedding strategies, chunk sizing math, vector DB (Postgres pgvector + Pinecone), hybrid search (BM25 + dense), rerankers, query rewriting. When RAG is the right answer and when it's NOT.
Agent loops + production AI
ReAct loop, planning vs executing splits, error-recovery, output validation as structured JSON, citations. Multi-agent orchestration with Anthropic's beta APIs.
Evals
Eval-driven dev for LLM apps: golden-set construction, LLM-as-judge with confidence calibration, regression testing prompts, A/B testing prompt versions in production.
Frontier topics
Vision models, prompt caching (90% cost reduction), compaction for long contexts, Files API, Skills, Batches API, voice agents with audio streaming, MCP servers.
Questions
Why AI engineering, not generic ML?
The highest-paying junior role in 2026 is AI engineering — wiring production AI into existing products. ML engineering (training models, MLOps) pays well but needs years. AI engineering builds on Python skills you already have.
Do I need ML/PhD background?
No. AI engineering is API-driven — you call Anthropic/OpenAI/Gemini, not train models. Solid Python + an understanding of HTTP + JSON is all you need. The math you'd learn in ML programs isn't relevant here.
Which model API do you teach against?
Anthropic's Claude is the primary teaching target (Messages API + tool use + caching + vision + voice). OpenAI Chat Completions is shown for comparison. Gemini gets a dedicated lesson for its free tier + long context window.
How much will the API calls cost while I learn?
Most lessons use mock responses (we ship deterministic stubs so you can test the loop). When real calls matter, the cheapest path is Gemini 2.5 Flash free tier (10K tokens/min). You can finish the track for under $5 total in real API spend.
Is this production-ready or playground stuff?
Production. Every module ends with a build-and-grade capstone using a pattern you'd ship at work. The customer-support agent capstone is structured exactly like the agents Anthropic publishes in their cookbook.
Do I need Python Foundations first?
If you're new to Python — yes, finish Python Foundations first (15 lessons are free). If you have 6+ months Python and have called any HTTP API before, start here.