The knowledge layer for risk and regulation — a reference you trust, models you can touch, and production-grade AI built with governance you can defend.
Three layers, one spine: understand the regulation, play with the model, then turn it into a production capability you can defend to a regulator.
A structured reference across ECB, EBA, RBI, Fed and BIS — COREP/FINREP, IFRS 9, Basel, IRRBB, FATCA — explained, not just linked.
Two kinds, playable in the browser: IFRS and regulatory computations, and prediction models — logistic, random forest, XGBoost, neural nets.
Production PD, LGD and forecast models — XGBoost, neural nets, transformers — shipped with model governance.
The backbone of the platform — IFRS/regulatory computations and prediction models across each domain, growing as we build.
Most show one side. We hold both — AI doing the risk work, and the AI itself governed as a risk.
Every model ships with a model card, validation evidence, a challenger, drift monitoring and a human approval gate.
Mapped to SR 11-7, the EU AI Act and Basel model-risk expectations — defensible to a validator on day one.
Purpose, data, method, metrics, approver — per version.
SHAP-based attribution on every prediction.
Monitored against a benchmark; retrain triggers.
A growing reference built from primary sources and working models, not summaries.
IRB, standardised, output floor, endgame.
Staging, PD/LGD/EAD, worked computations.
Templates, taxonomy, validation.
Cross-border and data obligations.
RegvedaAI is an independent, work-in-progress effort to make risk and regulation easier to understand and to build — explained clearly, simulated openly, and governed by design. It is built in the open and grows over time.
For risk and regulatory teams who want production-grade AI they can defend — not a slideware promise.