Changing role of engineers and productivity

Uber’s AI infra, from the inside

Topic 11 of 11

Published Oct 14, 2026 · Clip 1:21:22–1:31:51 (10:29)

Udai reflects on the changing engineering job: constantly removing the next bottleneck to maximize features shipped, and the 'capability overhang' of models not yet fully leveraged due to missing orchestration and tooling. On productivity, Uber is shifting from diff velocity to feature velocity, tagging PRs by category and using AI to cluster PRD-to-task tracing to measure P50 feature time by complexity. He describes how his own 'hands-on' definition changed—context engineering, giving right specs/constraints, multitasking across agent sessions, and rapid verification are the new core skills—and gives an example of automating customer support policy conversion from runbooks to agent-executable policies, cutting weeks to days. He closes recommending personal agents for automating personal work.