Gergely Orosz · 7 min read · Gen AI Tools
How Uber uses AI for development: inside look
Read the original on newsletter.pragmaticengineer.com →
Uber's session details its internal AI stack and tooling designed to scale AI across engineering, from multi-layer agent systems to developer-facing tools. It covers why the work was necessary, the resulting workflows, and the real-world impact and costs of making Uber a GenAI-powered company.
Key Takeaways
- Uber's internal AI stack uses four agentic layers: an internal platform, context sources, industry tools, and specialized agents for testing and code review.
- They built internal tooling (MCP Gateway, Uber Agent Builder, AIFX CLI) to simplify AI tool usage and improve agent effectiveness for developers.
- New developer tools like Code Inbox, uReview, Autocover, Shepherd, and a background agent platform called Minion enable scalable AI-generated code and large-scale migrations.
- Adoption challenges remain: top-down mandates are less effective than peer-driven wins and sharing best practices among engineers.
- Impact metrics show strong adoption and cost growth: 92% of devs use agents monthly, 31% of code is AI-authored, 11% of PRs are opened by agents, with AI-related costs up 6x since 2024 and a focus on token optimization.
Excerpt
Before we start: all The Pragmatic Summit videos are now available to view . Paid newsletter subscribers also have access to each session with the Q&A session, as well . I spent four years working at Uber until 2020 and experienced firsthand the company’s standout engineering culture. Uber is a company that did the speed run of going from a small startup, through hypergrowth, to being a large company facing major risk during the pandemic, when the rideshare business briefly collapsed. Today, it’s maturing as a publicly traded, profitable company, and employs almost 3,000 people in the tech function. At…