In active development
AutoScorecard
A credit scorecard automodeler with a pixel-art office interface: you lead a team of AI agents through chat.
Four AI agents turn raw data into a complete credit scorecard, along with a presentation deck.
- Agents: Partner, Prepper, Modeller, and Deck Creator do the work; you make the key decisions in chat.
- Scorecard engine: ChiMerge and WoE binning, variable selection (IV, correlation clustering, stepwise, VIF), logistic regression, PDO scaling, and validation with AUC, Gini, KS, and PSI.
- Data privacy: row-level data stays local; the agents only see metadata and aggregates.
Technology: Python · pandas · FastAPI · Pyodide · Cloudflare Workers · LLM agents
Open the app (opens in a new tab)
Restricted access, sign-in required.
