In this episode, Ben talks with Matas Rastenis, Senior Staff Engineer at Uber's Developer Platforms team, about what it actually looks like to bring AI tools to one of the world's largest engineering organizations  not as a flashy demo, but as production-grade infrastructure that delivers measurable impact. Matas shares how Uber approaches AI augmentation for its engineers: evaluating which tools to adopt, building the enablement infrastructure to support them, and making sure they're reliable and cost-effective at scale.

A key focus of the conversation is measurement. How do you actually prove that AI tools are making engineers more productive? Matas breaks down how Uber thinks about quantifying the value of AI-assisted development, moving beyond vanity metrics to real impact on developer velocity and output quality. They also dive deep into the organizational side — the change management challenge of rolling out new AI-powered workflows across thousands of engineers with different habits, skill levels, and levels of skepticism. Matas shares practical lessons on what works when driving adoption: meeting engineers where they are, building trust through reliability, and ensuring tools deliver value beyond slick demos. A must-listen for anyone leading developer experience, platform engineering, or AI adoption initiatives in large organizations.

Matas Rastenis is Staff Software Engineer on Uber’s Developer Platform team, building AI devtools that run reliably at scale. He leads AI testing automation and AI quality infrastructure, turning agentic workflows into production-grade systems that are measurable, cost-aware, and dependable. With a foundation in SRE and distributed systems, he focuses on making AI tooling observable, governable, and easy for engineers to trust.

Uber at https://www.uber.com/

Engineering at Uber: https://www.uber.com/en-US/blog/engineering/

Matas on LinkedIn: https://www.linkedin.com/in/matasr/

Here is the Stellar Work Newsletter: https://substack.com/@stellarwork