Monitor your production models as batch processes or inference servers. Detect the earliest signs of model or data drift and easily calculate historical model performance when ground truth is available. Continuously collect fresh training data from production.
Revela's highly consumable API provides the lowest latency on real-time metrics. Lose the clutter by displaying monitoring data on the dashboards and platforms you already use. Securely deploy on-prem or in the cloud with Kubernetes.
Choose from over 20 monitoring metrics across data drift, feature correlation, and model performance. Compatible with tabular and binary data for covering language and vision applications. Upload your own proprietary metrics through custom metrics.
Monitors automatically keep track of metric values and will alert when unusual events occurs. Set custom alerting thresholds with granularity at the model and metric levels. Individual events are aggregated smartly to reduce "alert fatigue".