# Polanyi Labs

> Polanyi Labs generates synthetic behavioral data for recommendation systems. Give it a website URL and AI shoppers, cast from NVIDIA Nemotron-Personas-USA, browse the live site in real cloud browsers. Every interaction (search, filter, impression, view, cart add, selection) is returned as a recommendation-system event log, so teams can train and evaluate recommenders before they have real user history. No existing user data is needed.

Polanyi Labs is built for DTC and e-commerce brands facing the recommender cold-start problem: too few sessions for collaborative filtering or sequential models to learn from. Shoppers are hard-pinned to the target domain and stop before entering personal or payment details. Every event is marked `is_synthetic: true` and defaults to `training_use: offline_sandbox_only`.

How it differs from other synthetic data tools: tabular synthesizers (SDV, MOSTLY AI, Gretel) learn from data you already have; recommender simulators (RecSim NG, RecoGym, KuaiSim) need a user model you define; public datasets (MovieLens, Amazon Reviews) aren't your catalog. Polanyi generates new behavior on your own site from just a URL.

Quickstart:

```python
pip install polanyi
from polanyi import Client
result = Client().simulate(site="https://your-store.com", n=3)  # needs POLANYI_API_KEY
```

Outputs: `personas.json`, `recsys_events.jsonl` (also Parquet), `summary.json`. New accounts get $10 in free credits; usage is billed per simulated shopper journey.

## Docs

- [Docs: SDK, REST API, and event schema](https://polanyilabs.com/docs): quickstart, `Client.simulate` parameters, every event field and event type
- [How it works](https://polanyilabs.com/how-it-works): population, goals, browsing, capture, and delivery into your recommender
- [FAQ](https://polanyilabs.com/faq): outputs, privacy, pricing, cold start, comparisons

## Guides

- [Synthetic data for recommendation systems: tools and methods](https://polanyilabs.com/synthetic-data-for-recommendation-systems): cold start, a comparison of approaches and tools, and how to use synthetic events in training

## Company

- [About the team](https://polanyilabs.com/team): Taha Moula (CEO), Ahmed Abdelrehim (CTO), Belhassen Bayar (technical advisor)
- [Get started](https://polanyilabs.com/get-started): API key or enterprise demo
- Contact: tahamoulaa@gmail.com