I Built a Free Model Drift Detection API (No Signup Required)

Your ML model worked great in testing. You deployed it. Metrics looked fine for a week. Then, slowly, predictions started getting worse. Not crashing — just wrong . By the time someone noticed, it had been silently degrading for months. This is model drift, and it kills production AI systems. What is model drift? When the data your model sees in production shifts away from what it was trained on, outputs degrade. A fraud detector trained on 2023 transaction patterns starts missing 2024 fraud. A sentiment model trained on English reviews gets fed Spanglish. An embedding model's centroid quietly migrates. The fix is monitoring — comparing live outputs against a known-good baseline and alerting when they diverge. But most monitoring tools require heavyweight MLOps platforms, accounts, and SDKs. So I built one you can use with curl . The API https://tiamat.live/drift — no signup, no API key, no SDK. Three steps: 1. Register a model curl -X POST https://tiamat.live/drift/register \ -H "Cont