Developing model catalog for ML developers

My Role

Lead designer

Design, research, prototyping

Result

Replaced fragmented tool experience with a centralized, consistent interface

Context

BIM (short for Business Integrity Modeling) is an internal platform used by ML engineers to manage the lifecycle of machine learning models. These models are critical in supporting Meta's Business Integrity (BI) systems, enabling detection of policy violations and malicious behaviors across the platform. The BIM platform enables model development, experimentation, approval, deployment, and monitoring.

User Challenge

As the complexity of ML use cases grew, engineers faced usability challenges and increasing overhead in managing their workflows. These usability challenges led to:

• Difficulty in tracking model status across teams
• Errors in version management
• Misalignment between training, approval, and deployment workflows

Approach

• User Research: Shadowed ML engineers and documented their current tools and workflows
• Tool Audit: Evaluated existing solutions for model registration, training, approval, deployment, and monitoring
• Information Architecture: Proposed a cohesive experience map that supported traceability and transparency across the entire ML model lifecycle
• Rapid Prototyping: Delivered wireframes and high-fidelity prototypes for key screens, validated with stakeholders

Design Solutions