Skip to content
All work

AI & Data

CV Job Match Agent

A 4-agent LangGraph pipeline that scores any CV against any job posting.

The problem

Candidates tailor the same CV to dozens of postings with no clear signal of how well it actually fits. I wanted a tool that returns a match score, the specific gaps, and what to change, for any CV and any job description.

How I built it

  1. 01Split the work across four specialised agents orchestrated as a LangGraph state graph, running on Groq-hosted LLMs for low latency.
  2. 02Added conditional routing so that if any agent fails, the pipeline stops immediately instead of passing bad state downstream.
  3. 03Logged every agent step (name, input summary, output summary, latency) to Supabase so each run can be debugged and evaluated after the fact.
  4. 04Deployed the FastAPI backend on Render, the Next.js frontend on Vercel and PostgreSQL on Supabase.

Outcome

  • Produces a match score, gap analysis and actionable recommendations end to end.
  • Full per-step observability for every pipeline run.
  • Runs entirely on free-tier infrastructure: $0 total hosting cost.

Next project

EmergencIA