Resume Judge
Scores resumes using semantic embeddings and retrieval-augmented generation (RAG).
Resume Score Analyzer is an end-to-end AI resume scoring workflow.
Upload PDF resumes → select skills → receive strict 0–100 skill scores.
Pipeline:
Store: Split resumes into chunks and embed into ChromaDB.
Retrieve: Fetch relevant excerpts using similarity search.
Judge: LLM outputs strict numeric scores.
Goal:
Avoid keyword matching.
Evaluate contextual experience differences such as:
“used React once” vs “led multiple React projects”.
Technologies:
- LangChain
- ChromaDB
- OpenAI embeddings (text-embedding-3-small)
- GPT-4o-mini scoring model
Workflow:
Upload → Chunk → Embed → Retrieve → Score → Display matrix results.
Docker ChromaDB runs on localhost:8000.
Server:
Express + Multer + LangChain.
Client:
Next.js interface displaying skill cards with color-coded scores.
Output example:
{ "resumeId": "1761205988038", "scores": { "React": 72, "AWS": 35, "SQL": 80 } }