---
title: "Under the Hood: Scaling Responsible AI at Uber"
description: "Uber describes a company-wide Responsible AI program designed to govern models developed across its many platforms and teams as AI becomes central to products and services. Its five pillars are visibility, explainability, governance, education, and adoption, implemented through a centralized Model Catalog, automated metadata, feature attribution, lifecycle controls, and employee learning resources. Model Cards provide searchable model descriptions, metrics, deployment details, and documentation, while Michelangelo integrates global, local, and deep-learning explanations using PFI, SHAP-based methods, TreeSHAP, and integrated gradients. Governance is shifted left into product and engineering reviews, with completion prompts before deployment, and adoption work brings existing models into scope through lifecycle-aware, technology-agnostic, context-sensitive controls. The program gives engineers and oversight teams shared visibility into model behavior and risk while aiming to support continued AI growth without losing accountability and trust."
---

# Under the Hood: Scaling Responsible AI at Uber

[Uber](https://yomu.fyi/company/uber) · Jun 9, 2026

**Type:** Explainer

## Summary

Uber describes a company-wide Responsible AI program designed to govern models developed across its many platforms and teams as AI becomes central to products and services. Its five pillars are visibility, explainability, governance, education, and adoption, implemented through a centralized Model Catalog, automated metadata, feature attribution, lifecycle controls, and employee learning resources. Model Cards provide searchable model descriptions, metrics, deployment details, and documentation, while Michelangelo integrates global, local, and deep-learning explanations using PFI, SHAP-based methods, TreeSHAP, and integrated gradients. Governance is shifted left into product and engineering reviews, with completion prompts before deployment, and adoption work brings existing models into scope through lifecycle-aware, technology-agnostic, context-sensitive controls. The program gives engineers and oversight teams shared visibility into model behavior and risk while aiming to support continued AI growth without losing accountability and trust.

## Context

The program responds to Uber’s use of AI across many platforms, teams, and workflows, including Michelangelo and teams embedding generative AI. Its stated goal is to provide visibility, explainability, and accountability while integrating responsible practices into model development without adding separate rules for every new system.

## Approach / What changed

Uber combines a centralized Model Catalog and standardized Model Cards with feature-attribution tooling in Michelangelo, shift-left governance checks, deployment gating, training, documentation, and adoption work for existing models. The operating model uses lifecycle-aware governance, adaptive technology-agnostic controls, contextual review, and system-level visibility across teams, repositories, and workflows.

## Takeaways

- Model Cards serve as the searchable single source of truth for model descriptions, performance and accuracy metrics, deployment details, and supporting documentation; some fields are populated from system-generated signals.
- Michelangelo supports global, local, and deep-learning explanations through PFI, SHAP-based methods including TreeSHAP, and integrated gradients, with results linked automatically to each Model Card.
- Uber applies governance from product and engineering planning through deployment, then uses lifecycle-aware classification and contextual review to bring existing models into scope without relying only on static rules.

## FAQ

### How does the Model Catalog reduce documentation effort for engineers?

The Model Catalog centralizes model metadata in standardized Model Cards and makes it searchable and filterable. Several fields are automatically populated from system-generated signals, reducing manual effort and overhead.

### Which techniques does Uber use to explain model predictions?

PFI provides global feature importance, while SHAP-based methods explain individual predictions and TreeSHAP is used for tree-based models. Integrated gradients are used for differentiable models such as neural networks.

### How are governance checks integrated into the ML development process?

Governance begins in product and engineering document reviews through PRDs and ERDs, following a shift-left approach. Users are prompted to complete Model Cards before deployment, creating an auditable gating mechanism tied to delivery.

**Tags:** [AI](https://yomu.fyi/topic/ai), [AI Governance](https://yomu.fyi/topic/ai-governance), [Machine Learning](https://yomu.fyi/topic/machine-learning), [Model Evaluation](https://yomu.fyi/topic/model-evaluation)

- Source: [Uber](https://www.uber.com/us/en/blog/scaling-responsible-ai)
- Source URL: https://www.uber.com/us/en/blog/scaling-responsible-ai
- Ingested by Yomu: 2026-09-15T18:14:20.518Z

[Read original post](https://www.uber.com/us/en/blog/scaling-responsible-ai)