Skip to content
← Back

6º semester · 2025-1 · Domrock

Auxia

Evaluating and comparing LLM answers, with RAG.

Role:DeveloperRepository ↗
  1. S

    Situation

    There were no standard criteria to rate LLM answers in sensitive contexts, such as supporting caregivers of people with Alzheimer's.

  2. T

    Task

    Build an app that shows two answers side by side, collects 1–5 scores with justification, enriches prompts through a vector DB and exports data for fine-tuning.

  3. A

    Action

    • ›Designed the FastAPI backend route and folder architecture.
    • ›Implemented prompt enrichment (RAG) with ChromaDB.
    • ›Built JWT/OAuth 2.0 authentication and data export.
    • ›Wrote unit tests with PyTest.
    • ›Built the login, header and user drawer in Vue.js and refined the Figma designs.
  4. R

    Result

    The project where I was most the technical reference: backend architecture, RAG and tests. Python and FastAPI at a teaching level.

Hard skills

  • FastAPI
  • Python
  • RAG
  • ChromaDB
  • MongoDB
  • PyTest
  • JWT/OAuth2
  • Vue.js
  • TypeScript

Soft skills

  • Proactively spotting blockers
  • Teaching concepts (RAG) to the team

Stack

  • Vue.js
  • TypeScript
  • Python
  • FastAPI
  • MongoDB
  • ChromaDB
  • Figma
Next projectTECHELP→