← All work

Independent · applied AI

AI Codebase Q&A

Ask a repository questions, get cited answers.

Year
2026
Role
Solo · in progress
Status
In progress
Type
Independent

Retrieval-augmented Q&A over a GitHub repository with inline citations and an evaluation set for answer faithfulness.

The problem

RAG is easy to demo and hard to make trustworthy. The goal is citations you can click and an eval set that catches regressions.

Approach

  • Chunk and embed a repository; retrieve with re-ranking before answering.
  • Force citations back to source lines; refuse when retrieval is weak.
  • Maintain an eval set scoring faithfulness and answer quality.

What it proves

  • A direct, high-signal answer to the AI-engineering interview question every team is asking.

Stack

App

Next.jsTypeScript

AI

EmbeddingsRetrieval + re-rankClaude API

Next

Real-time Collab App