Uber CTO: AI Now Anchors How Its Engineers Write Code

Uber is ramping up AI in how it writes and reviews code, with the company’s top engineering leader describing a sweeping shift in day-to-day software work. The push isn’t about replacing developers so much as reshaping their roles — and it’s already changing how new features are planned, built, and vetted across Uber’s platforms.

What Happened

Uber’s Chief Technology Officer for Mobility & Delivery, Praveen Neppalli Naga, has said the company has retooled much of its software development around AI, noting that more than 90% of Uber’s engineers use advanced coding tools such as Cursor and Claude Code. In practice, that means AI helps generate routine scaffolding, create tests, and modernize legacy code — with engineers guiding the work and reviewing results before changes land in production. Those comments, made public on July 31, 2025, underscored a strategic bet: AI isn’t a sidecar; it’s becoming part of the engine of engineering at Uber. More than 90% of engineers use AI tools, according to Naga’s remarks at the time.

How Uber Is Using AI

Inside Uber’s toolchain, the most visible example is uReview, an in‑house, multi‑stage AI system that augments human code review. The platform analyzes diffs at scale to spot potential bugs, security issues, and style problems before code ships. Uber says uReview now assesses the vast majority of changes in its repositories, and developers continue to rate its feedback as useful — a sign the system is adding signal rather than noise.

By the numbers, Uber reports that uReview analyzes over 90% of Uber’s weekly ~65,000 diffs, with engineers marking about three‑quarters of its comments as helpful and acting on most of them. Those adoption metrics — paired with Uber’s broader use of AI for test generation and routine edits — illustrate how AI is being embedded not just in coding, but in quality control and maintenance workflows as well.

Why It Matters

The shift points to an industry‑wide rebalancing of engineering work. If AI shouldered more of the boilerplate and review overhead, teams could spend more time on architecture, product tradeoffs, and reliability. At the same time, Uber’s leaders emphasize that human oversight remains essential: engineers still direct, verify, and approve changes, particularly for complex systems where context and judgment matter.

What’s Next

Uber says it plans to keep expanding AI’s remit in development — including deeper context for reviews and broader coverage areas like performance and test completeness — while keeping engineers firmly in the loop. The near‑term goal is pragmatic: fewer blind spots, faster feedback, and higher‑confidence releases at Uber’s scale.

Sources

• AI cracks coding: How new tools are already changing software jobs — Washington Post (July 31, 2025)
• uReview: Scalable, Trustworthy GenAI for Code Review at Uber — Uber Engineering Blog (August 12, 2025)
• Unlocking Financial Insights with Finch: Uber’s Conversational AI Data Agent — Uber Engineering Blog (July 17, 2025)