One of six curricula in the AI Career Curriculum
AI Engineering
You build with the model: you turn it into products and agents people use.
11 role tracks and 70 modules, free and MIT-licensed.
- AI WorkforceWork alongside the models
- AI StartupTurn it into a company
- AI GovernanceGovern, secure, and assure AI
- AI EngineeringBuild with the models
- ML EngineeringBuild and train the models
- AI InfrastructureRun the platforms models live on
Your relationship to the model
To you the model is a component, usually one you didn't train and can't change. You shape its behavior from the outside with prompts, context, retrieval, tools, and orchestration, and you're responsible for what it does inside your product.
This org covers AI engineering — building with the models: LLM application development, retrieval and RAG, agents and multi-agent systems, evaluation, guardrails, harness engineering, and production operations.
What you own
- Prompts, context, and retrieval that steer model behavior
- Tools, agents, and multi-step orchestration
- Product-level evaluation: does the feature actually work for users?
- Guardrails and fallbacks in the application layer
- Choosing which model to call for which job
What you don't own: Training the model, operating the serving fleet, or setting organizational policy. You work inside those constraints and report back when they don't fit the product.
The tracks
Ordered by level. Each one opens with Module 00, which maps the role's place relative to the model before the technical work starts.
Builder (L20)
LLM fundamentals, prompting & structured output, tool calling, retrieval basics, first agent, shipping
Engineer (L30)
Trace instrumentation, trajectory & tool evals, LLM-as-judge, RAG evals, eval-gated CI/CD, online evals
Agent loop, frameworks, RAG & memory, multi-agent, evals, guardrails, production, harness engineering
AI feature design, production prompting, RAG in product, orchestration, product agents, guardrails
Ingestion & chunking, embeddings, vector DBs & ANN tuning, learning-to-rank, hybrid retrieval, reranking
Customer discovery, integrating into enterprise stacks, evals on customer data, demo→production, field feedback loops
STT/TTS & speech-to-speech, latency & turn-taking, telephony/WebRTC, vision inputs, voice & multimodal evals
Senior & Architect (L40+)
Internal agent platforms: harnesses, tool & MCP registries, sandboxing, permissions, shared evals & tracing, paved roads
Agent systems from an architecture, eval & observability infra, multi-agent at scale, reliability, cost
Hiring & leveling AI roles, planning non-deterministic work, eval-driven delivery, vendor risk, teams that use coding agents
When not to use an agent, orchestration & memory architecture, eval harnesses, security, durable execution
Who you work with
No curriculum owns the model end to end. These are the handoffs with each of the others.
AI Workforce
Work alongside the models
- You hand them
- Tools, agents, and integrations that fit how people actually work, with sensible defaults and guardrails
- They hand you
- Workflow requirements and failure reports from daily use, which tell builders what to fix and what to build next
AI Startup
Turn it into a company
- You hand them
- Working features, honest demos, and a clear picture of what the product can promise customers today
- They hand you
- Prioritized customer problems, what's been promised, and what 'good enough to ship' means for this market
AI Governance
Govern, secure, and assure AI
- You hand them
- System descriptions, eval results, and guardrail designs for each AI feature, plus incident reports when something goes wrong
- They hand you
- Risk tiers, guardrail requirements, and red-team findings to fix before (and after) launch
ML Engineering
Build and train the models
- You hand them
- Real failure cases and task data from production, which tell them where a fine-tune or a different model would actually help
- They hand you
- Models, fine-tunes, and embeddings with documented strengths, failure modes, and benchmark results on relevant tasks
AI Infrastructure
Run the platforms models live on
- You hand them
- Traffic shapes, latency budgets, and model-mix requirements, so capacity and routing match how the product actually calls models
- They hand you
- Serving endpoints with known latency, throughput, rate limits, and cost per call, plus the telemetry to debug them
What you'll work with
- Model APIs
- Anthropic, OpenAI
- Agent frameworks
- LangGraph, CrewAI, AutoGen, LangChain, LlamaIndex
- Tools & protocols
- Model Context Protocol (MCP), Pydantic, FastAPI
- Coding-agent harnesses
- Claude Code, Codex CLI, Claude Agent SDK
- Retrieval
- pgvector, Chroma, Qdrant, BM25 hybrid search, reranking
- Production
- Temporal, Docker, Kubernetes, GitHub Actions
- Observability & evals
- OpenTelemetry, Langfuse, LangSmith, Phoenix, Ragas, Inspect
- Safety & governance
- OWASP LLM Top 10, NIST AI RMF, EU AI Act
How the tracks work
Learn
Each track is a sequence of modules: lecture chapters, hands-on exercises, quizzes, and projects. Start with Module 00 and work forward.
Check your work
The paired solutions repo has a reference walkthrough for every exercise. Attempt each one first; the gap between your answer and the reference is where the learning happens.
Built in the open
The content is written with AI assistance and under ongoing human review, so it can contain mistakes. Report anything wrong through the repo's Issues; fixes ship continuously.