README.md

---
title: StockMatch
emoji: 📈
colorFrom: indigo
colorTo: yellow
sdk: gradio
sdk_version: 5.49.1
python_version: '3.12'
app_file: app.py
pinned: false
license: mit
---

# StockMatch AI

Matching everyday savers to stocks that help protect against inflation.

Answer six questions, or click one of three preset investor profiles, and the app
retrieves matching stocks and writes a personalised explanation.

**How it works**

1. **Dataset** — 12,500 synthetic stock records generated with `Qwen2.5-1.5B-Instruct`,
   hosted at [Kogann/stockmatch-synthetic](https://huggingface.co/datasets/Kogann/stockmatch-synthetic)
2. **Retrieval** — numeric features are verbalised into text, embedded with
   `BAAI/bge-small-en-v1.5`, and searched with FAISS
3. **Filtering** — risk, budget, sector and market are enforced as hard constraints;
   semantic similarity ranks what remains
4. **Generation** — `SmolLM2-1.7B-Instruct` writes one explanatory sentence per stock

All figures shown are printed directly from the dataset, so the language model
cannot fabricate them.

⚠️ All data is synthetic and generated for an educational project. This is not
investment advice.