DriftAutoML

DriftAutoML is a web app (with optional CLI agent) that monitors production ML models for data and concept drift, then automatically triggers safe retraining and redeployment workflows. It connects to common data warehouses and feature stores, profiles incoming data, and compares it to training distributions. When drift or performance degradation is detected, it proposes retraining runs with an AutoML backend, generates a model card, and runs gated evaluations (holdout, backtesting, fairness checks, latency/cost). If the new model beats the current one under predefined policies, it can open a PR to your model repo or push a versioned artifact to your registry for rollout. This is an AI app + traditional app combination: the “AI” is the AutoML retraining and evaluation automation, while the product value is operational reliability. It’s aimed at teams that already have models in production but lack MLOps maturity.