Mona

Here are some examples of use cases actively monitored by Mona: Customer LTV prediction and targeted recommendations in e-commerce, fraud detection and
loan underwriting in financial services, language detection and voice recognition in audio, and property risk analysis in real estate. These use cases span across different tech stacks and models from support vector machines to random forests trees to deep learning.

• Data collection and aggregation. Customers can collect and aggregate performance data from all parts of your AI environment including training data, test data, and inference data pipelines. Customers can import data into Mona using a logging client (e.g. Python and Java), using a REST API endpoint, or directly from a data lake/warehouse. Data can be imported from multiple sources and asynchronously. Both batch and real-time use cases are supported. 
• Data abstraction, transformation, and segmentation. The product includes a powerful schema configuration language for defining how data is abstracted, segmented, and imported into Mona. For example, customers can define validation checks on imported data, numerical ranges for data bucketing, and metrics to include in the automatic anomaly detection search. This flexibility enables customers to track sophisticated combinations of business and technical indicators, such as business performance metrics for your AI system. See how to create metrics within Mona here
• Investigation. Users can leverage the dashboard visualization for exploring and investigating trends in data and model behavior. The product includes a unique noise reduction engine and root cause analysis features enabling quick issue resolution and high signal-to-noise ratio for alerts. 
• Automated actionable insights. Users can view auto-generated insights about anomalous behavior within specific segments of data regarding drifts, biases, sudden changes, outlier segments, distribution changes, and rule-based validations (e.g. thresholds). Insights can also be used more broadly for model performance evaluation such as for A/B testing new model versions. See insight overview here.
• Alerting and automation. Users can leverage multi-channel alerting (e.g. via email, Slack, MSFT Teams, Pagerduty) and also set up automated downstream workflow triggers (e.g., model retraining) via webhooks. See the alerting overview here.