---
title: TalentFit
emoji: 🎯
colorFrom: indigo
colorTo: blue
sdk: gradio
sdk_version: 6.13.0
app_file: app.py
pinned: false
license: mit
---

# TalentFit — Resume ↔ Job Match & Salary Intelligence

A decision-support assistant for recruiting. Given a **résumé** (pasted text or an
uploaded scan/image) and a **job description**, TalentFit returns a **match score**,
an **expected salary range**, and a **natural-language rationale with a gap analysis**.

It is **not** an automated hiring decision system — outputs are explained estimates
meant to assist a human reviewer.

## AI blocks (combined into one coherent app)

| Block | Contribution | Integration |
| --- | --- | --- |
| **NLP** (primary) | LLM extracts skills / experience / seniority from text; semantic match vs. job requirements | feeds structured features to ML; writes the final explanation |
| **ML Numeric** (primary) | Salary regression from structured + NLP-derived features; ≥2 models compared | consumes NLP features; its prediction feeds the NLP explanation |
| **Computer Vision** (bonus 3rd) | Classifies an uploaded résumé scan (résumé vs. other) and reads its text | gates the pipeline and produces the text input for NLP |

**Pipeline:** image → *(CV)* validate + extract text → *(NLP)* extract + match →
structured features → *(ML)* salary + score → *(NLP)* explanation & gaps.

## Run locally

```bash
python -m venv .venv
.venv\Scripts\activate        # Windows PowerShell
pip install -r requirements.txt
setx OPENAI_API_KEY "sk-..."   # or set as an environment variable for the session

# 1) Train (writes artifacts/)
python src/train.py

# 2) Launch the inference app
python app.py
```

Training and inference are **separated**: `src/train.py` produces everything in
`artifacts/`; `app.py` only loads artifacts and serves the UI.

## Data sources

All datasets are **new** (not the semester's Zurich-apartment or dog-breed data) and
publicly available. See the project documentation (`documentation.md`) for the full list,
sizes, and roles.

## Documentation

Full project documentation: [`documentation.md`](documentation.md).