How to Build a Hiring-Worthy RAG Engineer Portfolio Project in One Month
ashishllm · x · 2026-10-04
A step-by-step guide to building an advanced RAG portfolio project that stands up in interviews, with experiment-driven decision-making at every stage:
- Build your vectordb: choose pgvector for CRUD needs vs a dedicated vectordb for scalability.
- Engineer chunking: paragraph, overlap, document-layout-aware, or LLM-reasoning chunking; split pipelines for text+images+tables.
- Evaluate your retriever like an engineer: test semantic vs hybrid retrieval on an eval set with Precision, Recall, MRR, HitRate, Context Recall.
- Choose an LLM defensibly: answer in layers — cost vs accuracy tradeoffs first, then compare 3-4 models in the same price tier on Faithfulness, Groundness, Answer Relevancy via LLM-as-judge or human eval.
More from coding & agent
- Writer uses generate-critique-rewrite loop to fix AI errors instead of forbidding them — ivan_bezdomny · 2026-10-04
- In the agent era, bad interfaces turn UX costs into performance and cost problems — realJessyLin · 2026-10-04
- Dimillian's Rust game takes shape: ragdoll, dismemberment and magic systems via Astra — Dimillian · 2026-10-04
- Cursor agent ran a full week on GPT-5.2 to build a browser with 3M+ lines of Rust — gnukeith · 2026-10-04
- Gemini CLI PR fixes safeJsonStringify: shared references wrongly serialized as [Circular] — zero-dot7 · 2026-10-04
- Claude Code turns 1,321 exercises into an interactive galaxy for $190 in tokens — Stabbz · 2026-10-04