Minimal ML Monitoring
Self-hosted monitoring for ML inference that tracks confidence, latency and throughput and detects drift with real-time alerts
The product is self-hosted monitoring software for ML inference that tracks confidence, latency, throughput and output mix per model and detects data and feature-level drift to catch performance degradation before users are affected. It sells to developers, data teams and operations teams in B2B organizations that run ML models in production. It is delivered as a Kubernetes-ready Helm chart with Python and JVM SDKs, storing metrics in ClickHouse inside the customer's cluster and sending real-time alerts via Slack or email.
Key features
- Real-time ML inference monitoring
- Confidence latency throughput tracking
- Output mix per model per minute
- Feature-level drift detection
- Slack and email alert notifications
- ClickHouse-powered metrics storage
- Python and JVM SDKs
- One-line log ingest integration
- Real-time dashboards and history
- No social media activity within the last 30 days
GTM channels
- Docs
ICP
- Data analytics teams
- Software developers
- DevOps sre teams