This project explores loan performance data interactively — entirely in the browser using TensorFlow.js. It keeps data local, enabling rapid experimentation without server round trips or privacy risks.
Why client‑side
For prototyping credit features, speed of iteration and privacy are paramount. TensorFlow.js lets us load, transform, and model data directly in the page, so analysts and PMs can test ideas quickly.
What it does
- Loads CSV/Parquet locally (no upload)
- Feature engineering on the fly
- Simple models to guide intuition
- Charts and slice-and-dice UI for analysis
From here, we can harden features for production workflows or integrate server APIs when needed.