What is NannyML?
The NannyML Python library provides a robust and data-driven approach to monitoring the real-world performance of machine learning models, even in the absence of ground truth labels. By leveraging the confidence-based performance estimation (CBPE) algorithm, developed by the core contributors, NannyML can accurately estimate model performance and detect when and why changes occur. This enables data scientists and ML engineers to proactively maintain the health and reliability of their deployed models, ensuring they continue to deliver accurate and consistent results over time
Highlights
- Real-world model performance estimation without access to targets
- Confidence-based performance estimation (CBPE) algorithm for accurate and reliable monitoring
- Alerts on model performance changes and insights into the underlying causes
- Supports a wide range of model types and deployment scenarios
- Integrates seamlessly with existing ML workflows and pipelines
Platforms
- Web
Social