Fiddler Launches Powerful Set of Abilities for Machine Learning Lifecycles

Fiddler, a provider of Model Performance Management (MPM) solutions, is debuting a set of capabilities in analytics, diagnostics, and vector monitoring to help monitor, explain, analyze, and improve trustworthy models.

The company has also introduced a free 14-day trial and new pricing plans, making it easy for users to quickly realize the benefits of the Fiddler MPM platform and pay for only what they use.

According to the company, as organizations continue to run a greater number of models in production, it is growing increasingly important that data scientists and machine learning (ML) practitioners resolve model issues quickly, reduce operational costs, and increase business impact by improving model outcomes.

Fiddler is helping practitioners create a continuous feedback loop in ML lifecycles and gain contextual insights in a single, unified view.

Advanced analytics will surface actionable insights, making it easy to map model performance to business KPIs, according to the vendor.

Data Science and MLOps teams can help organizations save time, effort, and money through sharp prioritization while focusing on delivering better business outcomes. Fiddler’s new capabilities include:

  • Customizable Charts and Dashboards for Better Decision Making: A collection of shareable reports with monitoring insights, dashboards increase team collaboration and alignment between ML practitioners and business stakeholders. In a single custom chart, teams can track as many as six metric queries and up to twenty inputs and outputs for one or more models at a time and compare multiple models in real-time.
  • Diagnostics and Root Cause Analysis (RCA) for Quick Resolution: Users can surgically drill down into problem areas by performing RCA and pinpoint the underlying reason that caused a model to underperform or drift. Once the problem metric is identified, the user can drill down into feature impact and importance, like data quality or errors, in order to improve the model.
  • Alerts Dashboard for Issue Prioritization: Powerful and flexible alerts are fully customizable and accessible through a central dashboard. Users will have full control over the type of issues they want to be alerted on along with the timing for such alerts to reduce alert fatigue. Based on the severity of the alerts, MLOps teams can prioritize the most important issues across models in testing and production.
  • 3D UMAP Visualization for NLP and CV models to Diagnose Complex Drift: Data scientists can locate and identify complex data drift occurring in high dimensional space and visualize where and how the drift happened using the interactive 3D UMAP visualizer. A 3D UMAP reduces the number of dimensions, making analysis easier. For example, due to an uptick in orders on out-of-stock products, an online retailer can view the product images within the UMAP visualizer and notice that some products are mislabeled or misclassified, causing data drift.

“As organizations continue to launch more models into production, it becomes increasingly challenging to effectively monitor, diagnose, and improve those models. Else it creates ripple effects in the form of model degradation, bias, non-compliance, and consumer mistrust,” said Krishna Gade, co-founder and CEO of Fiddler. “Fiddler was created to make ML models easier to manage and trust. We want everyone to be able to use AI to its full potential, in a manner that is both fair and reliable, and the free trial will give anyone interested in achieving that goal to do so quickly.”

Fiddler is also now compliant with SOC2 Type II—which covers the trust service categories of security, confidentiality, and availability—and HIPAA. Organizations operating under HIPAA can now securely leverage Fiddler to offer better patient experiences and faster and more accurate medical diagnoses and reduce fraudulent health insurance claims.

For more information about this news, visit


Subscribe to Big Data Quarterly E-Edition