TL;DR
Get business pricing on networking and server gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
A Pragmatic Engineer interview with Uber Distinguished Engineer Uday Kiran Medisetty offers a look at the company’s developer infrastructure and AI tools. Uber uses a model gateway, shared agent settings and spending controls, while engineers still review every pull request.
The Pragmatic Engineer has published an interview with Uday Kiran Medisetty, a Distinguished Engineer at Uber, describing how the company’s internal developer infrastructure supports AI coding agents. The account details a model gateway, tools for moving settings between coding assistants, and spending controls, while stating that engineers still review all pull requests.
Medisetty’s interview describes a stack built around infrastructure Uber developed as its services and operational needs grew. The company’s internal tools include systems for identifying service ownership and managing workflows, as well as newer AI-focused tools that connect coding assistants to company models, code and engineering knowledge. The interview highlights DevPods, Code Inbox, Slate and Wisdom as examples of the tools engineers use, though the supplied report excerpt does not explain each one in detail.
Uber’s Model Gateway, introduced by its platform team in 2023, is designed to let AI features use different model providers rather than depend on one vendor. According to the report, it supports Anthropic and OpenAI models, as well as open models run by Uber. The gateway also provides a point for security checks, latency monitoring, spending records, audit logs and session traces.
The company also uses AIFX, a wrapper for coding-agent harnesses including Claude Code, Codex and OpenCode. The interview says it carries engineers’ settings, skills and plugins across those tools. Uber assigns engineers spending tiers for AI use; managers can approve additional spending when an employee reaches a limit. A plugin called CostCheck offers suggestions for managing costs across sessions.
How Uber Governs Coding Agents
The interview shows how AI-assisted development depends on more than access to a model. Uber’s approach links coding tools with security checks, usage tracking and shared configuration, giving the company ways to manage systems that may involve multiple providers and agent interfaces. That infrastructure can matter to engineers working across a large, interconnected codebase, where tool access and consistent settings affect day-to-day work.
There is also a practical trade-off between faster code production and the effort required to verify changes. The report says code review has become a bottleneck and describes tools intended to help engineers assess changes, including AI review, risk scores and visual evidence such as screenshots or videos when relevant. Those aids are not described as replacing human review: the report says engineers continue to review every pull request.
Uber’s spending tiers reflect another operational concern: allowing employees to use AI for suitable work without treating token consumption as unlimited by default. The company’s stated controls offer a model for how large engineering organizations may balance experimentation with oversight, although the interview does not provide independent measurements of productivity gains or review-time reductions.
As an affiliate, we earn on qualifying purchases.
Infrastructure Built for Uber’s Scale
The interview places Uber’s AI tooling within a longer history of building internal systems. Medisetty joined in 2015, when the company had two services, according to the report. By 2019 it had thousands. The source says Uber had about as many services in 2026 as in 2019, despite a business around ten times larger than in 2019; it does not specify a precise service count for either period.
Uber built tools when existing products could not meet its needs, the report says. Examples include M3DB, developed to handle a reported 500 million metrics per second, and Cadence, a durable workflow system. The report says Chronosphere grew from open-source M3DB and Temporal was built on open-source Cadence. It also describes city-level operations as a design requirement across systems such as deployments, observability and data infrastructure.
That history helps explain why the company’s AI work is presented as an extension of its developer platform, rather than a standalone chatbot rollout. A common gateway and reusable internal tools build on infrastructure designed to connect engineering work with Uber’s services and operating needs.
“All PRs are still reviewed by engineers.”
— The Pragmatic Engineer, summarizing Uday Kiran Medisetty’s interview
developer infrastructure management software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Productivity and Cost Measures
The supplied report excerpt does not include detailed results showing whether the tools have shortened development cycles, reduced review time or improved software quality. It also does not provide the methodology behind the reported roughly 50% reduction in token costs, including the comparison baseline and exact measurement window. That figure should be read as a claim relayed by the report, not as an independently verified result.
Details about how the tools work in practice remain limited in the provided material. It does not specify the safeguards applied to individual agent actions, how risk scores are calculated, or what proportion of engineering work uses agents. The excerpt also does not supply a publication date for the episode, so the timing of the described practices cannot be pinned down more precisely.
As an affiliate, we earn on qualifying purchases.
Further Details From the Interview
The source points readers to the episode transcript and says the discussion is available through YouTube, Apple and Spotify, with timestamps listed on the episode page. Those materials may provide fuller explanations of the internal tools and Medisetty’s account of how agent use and code review are changing.
For now, the report presents Uber’s infrastructure as an evolving system: it supports several model providers and coding harnesses, while retaining spending controls and engineer-led review. The supplied material does not announce a specific launch date or next product milestone.
cost management tools for AI development
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the news about Uber’s AI infrastructure?
The Pragmatic Engineer published an interview describing how Uber’s internal developer platform supports AI coding tools, model choice, usage monitoring and code review.
Does Uber use only one AI model provider?
No. The report says Uber’s Model Gateway supports models from Anthropic and OpenAI, as well as open models that Uber runs.
Are AI tools replacing human code review at Uber?
The report says engineers still review all pull requests. AI review and other signals are described as aids for evaluating code changes, not substitutes for that review.
How does Uber manage AI spending?
Uber assigns engineers spending tiers. According to the report, a manager can approve additional spending after an engineer reaches a limit, and CostCheck provides suggestions for managing usage.
Has Uber measured productivity gains from these tools?
The supplied report excerpt does not provide quantified productivity results. It relays a company-reported reduction of about 50% in token costs over “the last few months,” but gives no exact time window or comparison baseline.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
