LLM Zoomcamp
by DataTalks.Club
Ten weeks, completely free, and you finish with a RAG system you actually shipped.
Overview
LLM Zoomcamp is DataTalks.Club's free 10-week course on building a production-style AI assistant over your own knowledge base, taught by Alexey Grigorev with Will Russell of Kestra and AI engineer Timur Kamaliev. It is deliberately sequenced as an engineering course rather than a model-theory one. Module 1 builds a RAG pipeline on keyword search and then makes it agentic with function calling; Module 2 swaps in semantic search with embeddings using minsearch, sqlitesearch and PGVector; Module 3 covers AI and data orchestration with Kestra; a data-ingestion workshop builds dlt pipelines over filesystem and REST API sources into DuckDB with marimo dashboards; Module 4 covers offline and online evaluation of both retrieval quality and answer quality, including LLM-as-a-judge; Module 5 covers monitoring, user feedback and live dashboards; Module 6 covers best practices including LangChain, hybrid search and reranking; and Module 7 is a worked end-to-end fitness-assistant project. The capstone requires an original application with a searchable knowledge base, retrieval pipeline, evaluation process, user interface and monitoring. Module 1 takes about two weeks, the rest roughly a week each, and the final project three weeks. Lectures are pre-recorded, so the live cohort exists for graded homework, a leaderboard, peer review and a certificate; self-paced learners get every material free but no certificate. Everything runs on a normal laptop with no GPU and roughly $1-5 of API credit.
At a Glance
- Topic
- RAG
- Level
- Intermediate
- Format
- Course
- Cost
- Free
- Duration
- 10 weeks, ~5-15 h/week (self-paced any time; live cohort started 8 June 2026)
- Provider
- DataTalks.Club
- Hands-on
- Yes — code/exercises
- Certificate
- Available
What You’ll Learn
- ✓Build a RAG pipeline end to end, from ingestion through to a served answer
- ✓Add function calling so the retrieval step becomes genuinely agentic
- ✓Implement semantic and hybrid search with embeddings, PGVector and reranking
- ✓Evaluate retrieval and generation offline and online, including LLM-as-a-judge scoring
- ✓Monitor a live LLM application with user feedback and dashboards
- ✓Orchestrate ingestion and indexing pipelines with Kestra, dlt and DuckDB
- ✓Ship a capstone with knowledge base, retrieval, evaluation, UI and monitoring
Highlights
- •Genuinely free with no paid tier — videos, homework and code all live in a ~7k-star GitHub repo
- •Evaluation and monitoring each get their own module, which most free RAG courses skip entirely
- •Self-paced any time, or join the live cohort for graded homework, a leaderboard, peer review and a certificate
- •Runs on any laptop with no GPU and about $1-5 of OpenAI credit, so cost is not a barrier
- •Graduates consistently report it takes longer than advertised — budget 5-15 hours a week and expect to grind through the final modules
Who It’s For
Best For
- ✓Software engineers moving into AI engineering who want a portfolio RAG project
- ✓Data engineers who need the evaluation and monitoring half of LLM applications
- ✓Self-taught developers who learn better with deadlines, peer review and a leaderboard
- ✓Anyone who has read plenty of RAG blog posts but never shipped a working system
Prerequisites
- •Comfortable writing Python and working in the command line
- •Basic Docker familiarity, since services are run locally in containers
- •No prior ML or LLM experience required, and no GPU needed
FAQ
What is LLM Zoomcamp?
DataTalks.Club's free 10-week LLM engineering course, in which you build an AI assistant that answers questions over your own knowledge base. Aimed at Python developers moving into AI engineering who want a shipped project rather than theory. It covers agentic RAG, vector and hybrid search, orchestration, evaluation and monitoring, and finishes with a capstone you build from scratch — a searchable knowledge base, a retrieval pipeline, an evaluation process, a user interface and live monitoring.
Is LLM Zoomcamp free?
LLM Zoomcamp is free to access.
What level is LLM Zoomcamp for?
LLM Zoomcamp is aimed at a intermediate audience. Recommended background: Comfortable writing Python and working in the command line, Basic Docker familiarity, since services are run locally in containers, No prior ML or LLM experience required, and no GPU needed.
How long does LLM Zoomcamp take?
Expect roughly 10 weeks, ~5-15 h/week (self-paced any time; live cohort started 8 June 2026). Most learners work through it at their own pace.
What will I learn from LLM Zoomcamp?
You'll learn: Build a RAG pipeline end to end, from ingestion through to a served answer; Add function calling so the retrieval step becomes genuinely agentic; Implement semantic and hybrid search with embeddings, PGVector and reranking; Evaluate retrieval and generation offline and online, including LLM-as-a-judge scoring; Monitor a live LLM application with user feedback and dashboards; Orchestrate ingestion and indexing pipelines with Kestra, dlt and DuckDB; Ship a capstone with knowledge base, retrieval, evaluation, UI and monitoring.
Topics
Sources
This page was written from 3 sources, 1 on domains other than github.com.