Module 0115–18 h
LLM Foundations and APIs
Transformers, tokenization, generation, current models, and multi-provider clients.
A multi-provider client with observable tokens, latency, errors, and metadata.
LIVING CURRICULUM · MAINTAINED BY PRACTITIONERS
The LLM Engineering course that will never be finished. Because AI won't be either.
Working practitioners spot what changed, debate it in public pull requests, and turn it into reviewed lessons, labs, and tools. You learn from that living edition; when you improve it, your name stays.
No account
Progress works locally. We do not store your code, prompts, or traces.
Commitment250–300 h
CurationPractitioners + PRs
Updated08/21/2026
Open source116 reviewable resources
FROM CHANGE TO LESSON
What one practitioner learns today, the whole community studies tomorrow.
The curriculum does not wait for a new edition. Every improvement keeps the discussion, sources, review, and authorship that made it trustworthy.
Active edition · 116 resources · reviewed 08/21/2026
- 01
The industry changes
New models, APIs, patterns, and risks appear before the course can cover them.
- 02
Someone spots the gap
A practitioner proposes a traceable improvement in the public repository.
- 03
Experts curate it
The community checks sources, runs examples, and debates the change.
- 04
The site learns
The approved improvement reaches the course and its author remains permanently linked.
PROFESSIONAL PROGRAM
A path that moves with the industry, not a pile of tutorials
Every module ends in verifiable evidence. Learn the theory you need, then turn it into systems you can explain in an interview or defend in production.
Module 0238–45 h
Advanced Prompt Engineering
Prompts as software: tools, structured outputs, evaluation, versioning, and mitigation.
A prompt-evaluation pipeline with a dataset, A/B tests, and regression gates in CI.
Module 0350–60 h
RAG Systems and Evaluation
Ingestion, embeddings, vector stores, chunking, reranking, grounding, and RAGAS.
A RAG application with citations, abstention, and automated evaluation.
Module 0450–60 h
AI Agents and Orchestration
Agentic patterns, LangGraph, MCP, tool use, multi-agent systems, memory, and evaluation.
A multi-agent system with an MCP server, traces, and failure recovery.
Module 0550–60 h
LLMOps, Production, and Responsible AI
Observability, continuous evaluation, cost, deployment, security, safety, and governance.
A deployed system with SLOs, costs, evaluations, and an operational runbook.
Module 0647–57 h
Production Capstone
End-to-end project, technical portfolio, deployment, defense, and final preparation.
An interview-ready system with architecture, metrics, costs, and a live demo.
IN-BROWSER LAB
Practice engineering decisions
Tools that produce useful artifacts: contracts, metrics, threat models, and operating plans. Your data is processed on-device.
01Design
Prompt Contract Studio
A versionable prompt and a test list before any model is called.
Open lab02DesignStructured Output Validator
An exact contract diagnosis and negative test cases.
Open lab03RAGChunking Workbench
An explainable chunking strategy with statistics and inspectable fragments.
Open lab04RAGRetrieval Metrics Lab
A report that separates retrieval failure from generation failure.
Open lab05EvaluationEvaluation Spec Builder
A JSON specification ready to become a regression suite.
Open lab06AgentsTool Schema Builder
A portable definition and a safe-execution checklist.
Open labPROFESSIONAL OUTCOME
Your learning ends in evidence, not a 100% progress bar
Build a technical portfolio with evals, ADRs, metrics, traces, and a defensible capstone. The passport summarizes your work; it does not pretend to be an official certification.
Build my portfolio- 01Deployable capstone with architecture and runbook
- 02Reproducible evals and negative cases
- 03Downloadable competency passport
- 04Preparation aligned with official blueprints
OPTIONAL PREPARATION
If you need an official exam, you can prepare for that too
It is one supporting tool inside the path: mock exams, flashcards, and domain-level repair versioned against official guides.
Verified blueprint · 2026-08-21
Prepare for AIF-C01AWS Certified AI Practitioner
AIF-C01 · 65 questions
Verified blueprint · 2026-08-21
Prepare for NCA-GENLNVIDIA-Certified Associate: Generative AI LLMs
NCA-GENL · 50 questions
FROM LEARNER TO CONTRIBUTOR
When you know something the course does not, you are already part of the club.
Open the pull request. If it improves the course and passes review, the site changes and your contribution remains linked to your profile. Recognition rewards useful knowledge, not lines of code.