AI & Data Engineering

Production-grade guidance for data pipelines, ML infrastructure, modern data stack engineering, prompt engineering, and LLM evaluation.

Start Here

1. Load only the reference files relevant to the current task (see Workflow Router below).
2. For data engineering: choose the processing model first — batch vs. streaming, Lambda vs. Kappa, warehouse vs. lakehouse.
3. For prompt/LLM work: choose the right pattern first — see Prompt Engineering Quick Reference below.
4. Use the scripts in scripts/ for common operations — don't write boilerplate from scratch.
5. Apply DataOps practices by default: data contracts, quality checks, CI/CD gates, and observability.

Core Rules