What it does

This repo holds the lab materials for a DataExpert.io course on shoehorning LLMs into data engineering workflows. Over four days, students progress from generating SQL queries with LLMs, to orchestrating pipelines with LangChain, to deploying a “ZachGPT” RAG chatbot via Pinecone. It’s essentially a curated walkthrough with video lectures, Python setup via uv, and a PostgreSQL dump of sample data.

The interesting bit

The course treats LLMs less as oracles and more as overeager interns: useful for drafting SQL from dimensional schemas, auto-generating LinkedIn posts in a specific human’s voice, and fielding business questions via retrieval-augmented generation. The progression from raw prompting to LangChain to vector databases mirrors how many teams actually adopt this stuff—incrementally, with growing API bills.

Key highlights

• Four structured days: prompt-to-SQL, LangChain orchestration, business-value automation, and RAG with Pinecone
• Uses a real PostgreSQL dump (halo_data_dump.dump) for hands-on query generation labs
• Requires OpenAI API key throughout; Pinecone key added for Day 4 vector search
• Setup via uv sync or fallback pip install .
• Links to external repos for Day 3 (auto-feedback) and Day 4 (vector database) labs

Caveats

• README is essentially a syllabus with setup instructions; no actual code or notebooks visible in the provided source
• Live cohort dependency: Day 2 lab mentions a cloud database URL “Zach gives in Zoom”—self-serve learners must restore the local dump instead
• Mac-centric PostgreSQL download link; Windows users get a terse “Or set it in Windows” for env vars

Verdict

Worth bookmarking if you’re a data engineer who learns best from guided, project-based courses and doesn’t mind paying OpenAI for the privilege. Skip it if you want standalone, runnable code without video homework.