Uday Kiran Medisetty is a Distinguished Engineer at Uber building the infrastructure for an agent-driven software factory. His work brings identity controls, privacy protections, organizational knowledge, and reusable tools into AI systems that write and maintain software across a large engineering organization.
Before turning to autonomous software development, Medisetty helped build foundational systems behind Uber’s transportation and delivery businesses. A project originating in a 2016 hackathon contributed to a real-time push platform that replaced repeated polling with bidirectional streaming between mobile applications and backend services. He also coauthored Uber’s account of Edge Gateway, its shared platform for configuring, managing, and monitoring APIs.
By 2020, Medisetty was a principal engineer at Uber. He subsequently helped lead a ground-up redesign of its fulfillment platform, adapting mission-critical marketplace architecture to multiple business lines. In 2026, he coauthored research on infrastructure failover that describes allocating recovery capacity according to business criticality.
Building autonomous development at Uber
Medisetty’s software-factory architecture adapts those distributed-systems concerns to enterprise AI:
- Governed model access: A centralized gateway authenticates workloads, redacts personal information, applies safety policies, tracks spending, and attributes requests to individual projects, teams, and users. Medisetty reported that its safeguards operate within 100 milliseconds and that the gateway supports more than 800 internal projects.
- Token-efficient agent tooling: An MCP Gateway and OmniMCP connect agents to internal APIs and external services through a common entry point. Tool discovery, command-line access, and generated scripts prevent large catalogs and responses from consuming excessive model context; Medisetty reported fleetwide token savings exceeding 40 percent.
- Reusable execution environments and skills: Uber adapted its existing remote-development infrastructure into isolated, pre-provisioned agent workspaces spanning large repositories. A managed marketplace adds discovery, automated quality checks, and evaluation feedback for reusable agent skills.
- Organizational context graphs: A shared graph connects service ownership, dependencies, code, design documents, incidents, and operational systems. Uber’s internal assistant, Cortana, makes that information and its associated tools accessible through Slack, web, and command-line interfaces.
Medisetty also championed adoption of uReview, Uber’s AI-assisted code-review system. His ambitions extend beyond engineering: he described embedding 30 engineers across 16 business functions and envisions agentic systems covering planning, implementation, validation, deployment, observability, and maintenance.
