Project brief — Application

Document Q&A Application

Build a service that ingests a document corpus, indexes it, and answers natural-language questions with citations pointing back to the source passage.

Intermediate
Type
Application
Difficulty
Intermediate
Duration
3 weeks
Milestones
3

Requirements

Non-negotiable. A submission that skips one of these is returned, not graded.

  • R1Chunking strategy justified in writing, not chosen by default
  • R2Retrieval evaluated against a hand-labelled question set
  • R3Every answer carries a verifiable citation
  • R4Graceful behaviour when the corpus does not contain the answer

Timeline

3 weeks, staged

  1. Week 1

    Ingestion

    Parsing, chunking and embedding with a reproducible index build.

  2. Week 2

    Retrieval & answering

    Query pipeline, citation binding and refusal behaviour.

  3. Week 3

    Evaluation

    Labelled question set, retrieval metrics and failure analysis.

Stack

PythonFastAPIVector DB

Deliverables

  • Running API
  • Evaluation report
  • Decision log
  • README with setup

Reviewed on

  • Retrieval quality
  • Citation integrity
  • Code clarity
  • Written reasoning

Take this brief

Every brief is reviewed by a practitioner. You get written feedback on code, documentation and the decisions behind both.

Request a review slot