Staff Software Engineer - Autonomous Verification Platform

About the Role and Team

The Test and Verification team is part of Developer Platform, whose mission is to empower engineers at Uber to build reliable, scalable software efficiently and without friction. Developer Platform builds the languages, tools and infrastructure that thousands of Uber engineers, and increasingly AI agents, use every day to write, test and ship code. Within that, our team builds the systems that determine whether a change is safe to land. We focus on backend services at Uber's scale and work closely with the teams building agents and automation. Our goal is to make verification fast, trustworthy and meaningful, so Uber can ship more with less manual review.

Join us to help solve the biggest challenge in modern software development: verification. Code generation is cheap, so the bottleneck is knowing whether a change is correct and which tests actually tell us. In an agentic world, the question is no longer whether code was reviewed, but whether it was verified. We are looking for someone with vision to define what trustworthy verification looks like when agents write and check much of our code.

In this role you will build the systems that decide whether backend changes are safe to land. You will work hands-on to give humans and AI agents a clear signal on which tests are meaningful and trustworthy, and you will help set the team's technical direction. You will be a key contributor to the verification layer of Uber's software factory.

What You’ll Do

• Build systems that measure the quality of backend tests beyond coverage, using techniques such as fault detection, mutation analysis and replay of real historical bugs.
• Design and build the signals that tell engineers and agents which tests matter for a change, which are reliable, and what a failure means.
• Build verification for AI-generated code, including independent test generation, behavioral validation and verification artifacts that CI can attest to.
• Define how we evaluate agents, skills and validators that write or check code, using golden datasets, replay of real bugs and regression tracking across models and prompts.
• Measure whether our verification signals can be trusted, such as how often judges and validators agree with real outcomes, so humans and agents know when to rely on them.
• Reduce flakiness and make failures actionable, so engineers and agents know what failed, how to fix it and who to contact.
• Define and track verification health, such as defect catch rate and flake rate, and catch regressions early.
• Shape the vision for verification in an agentic world: spec-driven development, independent test generation, and verification evidence that CI can attest to, so human judgment moves from reviewing every PR to defining what correct means.
• Contribute to the team's technical direction: write design docs, set standards, mentor engineers and partner with CI, build and product teams.