Extend Amazon SageMaker with Fiddler to improve model performance

Train and deploy models with Amazon SageMaker or other ML pipelines and monitor, explain, analyze, and improve models with Fiddler.

A single pane of glass for all models

Command center-like visibility into the behavior and performance of all models in training and production.

Purpose built for the ML lifecycle

The Fiddler MPM platform is a pure-play solution with a focus on model monitoring, explainable AI, analytics, and fairness. 

Use Fiddler to monitor model behavior, explain the how and why behind model degradation, analyze the root cause, and deliver fair outcomes.

Customer driven

Our customers inspire our product roadmap and we aim to deliver product enhancements in a nimble and agile fashion.

Our platform saves engineering time and costs so that you can focus on domain and application-specific improvements.

Deep expertise and care

Our Data Science and Customer Success teams care deeply about your onboarding and model deployment and improvement.

They are like an extension of your team and work closely to ensure improved model predictions that deliver business value.

An integrated solution for reducing time to detection and resolution

Fiddler attaches as a sidecar to your ML pipeline and integrates into your current ML workflows to make them richer and more complete.

Fiddler + Amazon SageMaker partner integration
Fiddler AI monitor data integrity

Model Monitoring

  • Detect performance issues including drift, bias, and integration errors
  • Receive real-time alerts on performance conditions and slices
  • Log and analyze model traffic, metadata and versions

Explain and Analyze

  • Ingest models and generate human-readable explanations from complex multimodal deep learning models
  • Slice datasets in real-time to compare metrics and validate predictions
  • Root-cause performance issues with explainability
  • Evaluate and connect model metrics to business goals
Fiddler AI measure fairness metrics

Assess Fairness

  • Assess biases, including hidden intersectional biases, against practical fairness metrics on both training and production datasets
  • Simulate different cases with counterfactual explanations
Fiddler AI monitor data integrity

Model monitoring

  • Detect performance issues including model drift, model bias and integration errors
  • Receive real-time alerts on performance conditions and slices
  • Log and analyze model traffic, metadata and versions

Explain and analyze

  • Ingest models and generate human-readable explanations from complex multimodal deep learning models
  • Slice datasets in real-time to compare metrics and validate predictions
  • Root-cause performance issues with explainable AI
  • Evaluate and connect model metrics to business goals
Fiddler AI feature attribution
Fiddler AI measure fairness metrics

Assess fairness

Get started

Fiddler is a SageMaker Highlight Partner and is listed on the AWS Marketplace for easy click-through procurement and AWS cloud billing. You can contact our sales team for help with the process.