RADIX Talent Match

An end-to-end tool that tells a candidate how ready they are for a specific job, built on RADIX's 12-skillset framework. Upload a JD or résumé, build a profile, and run Talent Check (vs a company's bar) and Skill Match (vs one JD).

All 5 hackathon modules, one integrated app:

#	Module	Where it lives
1	JD Analytics	backend/app/extract.py → analyze_jd() (LLM)
2	Résumé Parsing	backend/app/extract.py → analyze_resume() (LLM)
3	Profile Builder	frontend/src/components/ProfileView.jsx + Supabase
4	Talent Check	backend/app/match.py → talent_check()
5	Skill Matching	backend/app/match.py → skill_match()

Architecture

React (Vite)  ──HTTP──▶  FastAPI (Python)        Supabase
:5173                    :8099                    (auth · DB · storage)
  │                        │                         ▲
  │  parse / match calls   │  Groq LLM (extraction)  │  profiles, JDs,
  └────────────────────────┘                         │  company bars, history
        │                                             │
        └──────────── auth / data / files ────────────┘

• LLM is provider-agnostic (backend/app/llm.py): Groq (default), OpenAI, Grok, Anthropic, or a local fallback — switch with LLM_PROVIDER.
• Scoring is plain, explainable Python. The "what is ready" knobs live at the top of match.py.

Prerequisites

• Python 3.11+ and Node 18+
• A free Groq API key (https://console.groq.com/keys) — or set LLM_PROVIDER=local
• A Supabase project (free tier) — only needed for login/saving; the demo runs without it

Setup

1. Backend

cd backend
python -m venv .venv
.venv/Scripts/python -m pip install -r requirements.txt   # Windows
# cp .env.example .env  and fill in GROQ_API_KEY  (already set for this repo)

2. Supabase (optional, for auth) — see supabase/README_SUPABASE.md

• Run supabase/migration.sql in the SQL editor.
• Put your project URL + anon key in frontend/.env.
• Turn OFF email confirmation for smooth demo signups.

3. Frontend

cd frontend
npm install
# cp .env.example .env  and fill in VITE_SUPABASE_* (already set for this repo)

Run (two terminals)

# terminal 1 — backend
cd backend && .venv/Scripts/python -m uvicorn app.main:app --port 8099

# terminal 2 — frontend
cd frontend && npm run dev

Open http://localhost:5173. The dashboard works immediately in demo mode.

Ports: backend 8099, frontend 5173. (8000/8077 were occupied on the dev machine.)

Regenerate the baked demo data

cd backend && .venv/Scripts/python make_demo_data.py    # writes frontend/public/demo_data.json

5-minute demo script

1. Open the dashboard — "AI engine live" banner shows Groq is wired. Switch the Viewing as dropdown across the 4 sample candidates: cards, gauges, and the Talent Check panel all recompute. Ananya (systems) tops out on Google SWE 88%; Karthik (data) on Microsoft Data Analyst 86% — scores are not constant.
2. Build a profile (My Profile) — upload a résumé PDF; AI auto-fills skills. Sign in, click Save.
3. Talent Check — pick a company → readiness score + per-skillset gap bars.
4. Skill Match — pick a JD → match score + matched/missing skill lists.
5. Employer (Hire) — upload a JD; AI extracts required skills; post it.
6. Admin — edit a company's 12-skillset bars; re-run Talent Check to show the score move (proves the bars are live, not hardcoded).

Test the engine standalone

cd backend && .venv/Scripts/python test_engine.py

Prints the full JD/résumé extraction + the Talent Check / Skill Match matrix.