Skip to content

HawkEye

Debug Meta machine learning workflows end to end

Visit official site

What HawkEye is

The toolkit is described as an internal Meta tool for monitoring, observability and debugging of machine learning workflows.

What you can do with it

  • Collect service, training model and data generation data
  • Analyze and detect models with dropping prediction quality
  • Isolate feature subsets through model explanation
  • Compare snapshots and visualize a model graph

Application scenarios

  • Link a metric anomaly to a specific model snapshot
  • Find a related feature subset when predictions shift
  • Discover label imbalance from training data statistics

Pros & cons

✓ What we like

  • Guided UX workflows support exploration, investigation and mitigation
  • Current snapshots can be compared with earlier stable ones

! What to watch out for

  • The material describes it as an internal Meta toolkit
  • Analysis depends on collected service, model and data generation data

FAQ

What is it used for?

The material lists monitoring, observability and machine learning workflow debugging.

Which models are named?

The material names recommendation and ranking models for many Meta products.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review