Sumanth Sukumar

Building Products in the Age of AI: Inside Uber’s Software Factory

AI is changing how teams at Uber build products, reshaping every step from the first idea to what customers experience. In this session, we'll share how Uber built its Software Factory: a lifecycle-wide system of agents, infrastructure, and workflows that spans Plan → Build → Validate → Deploy → Observe → Maintain, and what we're learning as these new ways of working evolve at scale. We'll walk through how work moves through the lifecycle today, where agents take on tasks that once required a handoff between product, design, and engineering, and how that is compressing the loop between understanding a problem, shaping an idea, prototyping it, shipping it, and learning from customers. A growing share of that work is no longer initiated by a person at all: managed agents review code, heal failing builds, complete end-to-end changes with visual validation, triage alerts, and handle maintenance, with humans reviewing and escalating. Today, more than 70% of Uber's pull requests originate from local or cloud agents, and teams have built thousands of reusable agent skills across the lifecycle. We'll also look at the technology behind this shift, including how we enable agents to securely understand Uber's systems through a company-wide context graph, gateways for models and tools, and cloud development environments, and how we measure whether each agent is delivering value, not just output. Along the way, we'll share what changed, the challenges we hit, and where human creativity, context, and judgment remain essential.