A free bootcamp that actually builds to production

What it does

LLM Zoomcamp is a free, open-source curriculum that walks you from basic Python to a production-ready RAG application. The repo holds pre-recorded lectures, Jupyter notebooks, homework, and a capstone project template. You build a knowledge-base Q&A system, add vector and keyword search, wire in evaluation and monitoring, and end with a deployed assistant you actually own.

The interesting bit

The course is deliberately stingy with prerequisites — “if you can write a Python function and have heard of ChatGPT, you have enough.” No GPU, no prior ML, and roughly $1–5 in API credits. The 2026 syllabus starts with agentic RAG rather than treating it as an advanced extra, which suggests the instructors think the old “basics first, agents later” sequencing is backwards for practitioners.

Key highlights

• Runs fully on a laptop; cloud APIs handle the heavy lifting
• Covers the unglamorous stuff: hybrid search, reranking, offline/online evaluation, monitoring dashboards
• Uses Kestra for orchestration and dlt for data ingestion — real tools, not toy wrappers
• Peer-reviewed capstone with past projects (fitness bots, medical FAQ assistants, codebase Q&A) as reference
• Self-paced mode is always open; live cohorts add deadlines, scoring, and a certificate

Caveats

• “Live cohort” is a misnomer: all lectures are pre-recorded, so the value is community pressure, not interactivity
• Homework forms close after deadlines; late joiners to the live track may miss graded submissions
• Certificate requires peer-reviewing three other projects, which is fair but adds friction

Verdict

Worth bookmarking if you’re a software or data engineer who needs to ship a RAG system without wading through blog-post tutorials. Skip it if you want deep LLM theory or fine-tuning; this is plumbing, not research.