TL;DR
Get business pricing on monitors, keyboards and dev 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 describes how the company is connecting AI agents to its developer systems and managing model choice, security and costs. Uber says engineers still review every pull request, while AI tools provide additional review signals and supporting context.
The Pragmatic Engineer has published an interview with Uday Kiran Medisetty, a Distinguished Engineer at Uber, describing how the company’s internal developer infrastructure is being adapted for AI agents. The account details systems for routing model requests, carrying settings across coding tools, controlling usage costs and assisting code review, offering a view into how a large engineering organization is integrating AI into existing workflows.
Medisetty, who leads much of Uber’s work on agentic AI according to the report, walks through both older infrastructure and newer AI-focused tools. Uber’s Model Gateway, introduced by the platform team in 2023, is designed to let internal systems work across vendors including Anthropic and OpenAI, as well as models Uber runs itself. The gateway also supports security checks, spending records, audit logs and session traces, the report says.
Uber’s AIFX layer wraps coding harnesses including Claude Code, Codex and OpenCode. The stated purpose is to let engineers change harnesses without losing settings, skills or plugins. The report also describes background agents and internal systems such as Slate, Code Inbox and DevPods, though the supplied material does not give a full technical account of each tool.
Spending is managed through allocated usage tiers, with managers able to approve additional monthly spend when an engineer reaches a limit. A plugin called CostCheck offers suggestions across sessions. The report says Uber reduced token costs by about 50% in recent months, but the provided material does not specify the measurement method or exact period. On code review, Medisetty says engineers still review all pull requests; AI review, risk scores and supporting material such as screenshots or videos are intended to help reviewers judge the work.
How Uber Governs AI Coding
The account illustrates that deploying coding agents inside a large company involves more than choosing a model. Uber’s approach, as described, combines a shared model gateway with security checks, usage tracking and tools that make different coding harnesses easier to manage. That structure can give the company flexibility to change providers while maintaining oversight of requests and costs.
Code review remains a human responsibility in the report, even as AI contributes assessments and context. This distinction matters to engineering teams weighing automation: the tools may help prioritize review effort, but the source does not say they replace a reviewer’s decision. The reported cost reduction also highlights that usage controls are part of the operational design, rather than a separate budgeting issue.
AI code review tools for developers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
From Internal Tools to AI Agents
Uber’s AI systems build on a long history of internal engineering tools. The report says the company had two services when Medisetty joined in 2015, and thousands by 2019. It estimates the service count in 2026 will be around the same as in 2019, despite the business being about ten times larger than it was then. These figures are presented in the interview, not independently validated in the supplied source.
The report attributes some of Uber’s in-house tooling to gaps in available commercial products and to the company’s scale. Examples include uOwn, for identifying service ownership; M3DB, for handling metrics; Cadence, a durable workflow engine; and Morpheus, which combines feature flags and experimentation. The company’s need to represent cities across systems—including deployment, observability and infrastructure—also shaped its tooling. That existing platform is the setting in which Uber is now connecting agents to code, services and engineering knowledge.
“All PRs are still reviewed by engineers.”
— Uday Kiran Medisetty, as described in the report
As an affiliate, we earn on qualifying purchases.
Limits of the Reported Results
The supplied report material does not provide technical benchmarks for the gateway or agent tools, nor does it detail how the reported 50% token-cost reduction was calculated, what baseline was used or which workloads were included. It also does not quantify changes in engineering productivity, review times or defect rates.
Some of the systems are named without detailed explanations of their operation, and the account does not specify how many engineers use each tool or how broadly the AI workflow has been deployed. The material describes Uber’s approach through an interview; it is not an independent evaluation of the systems’ performance or security.
AI security and audit logs software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Further Detail on Agent Workflows
The episode is presented as a tour of Uber’s developer tools and AI infrastructure, with a transcript and episode timestamps referenced by the publisher. The supplied material does not announce a product launch, rollout date or new company-wide policy. Further reporting or detail from Uber would be needed to establish deployment scale, productivity outcomes and the methods behind its cost figures.
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 with Uber Distinguished Engineer Uday Kiran Medisetty describing internal tools used to connect AI models and coding agents with engineering workflows.
What does Uber’s Model Gateway do?
According to the report, the gateway supports multiple model providers and models run by Uber. It also connects to security checks, spending records, audit logs and session traces.
Does AI now approve Uber’s pull requests?
No such change is described. The report says engineers still review all pull requests; AI review and other signals are used to support that process.
How does Uber manage AI costs?
The report describes allocated usage tiers, manager approval for additional monthly spend and a CostCheck plugin. It reports a roughly 50% reduction in token costs in recent months, without specifying the baseline or calculation method in the supplied material.
What is not yet known about the tools?
The supplied account does not quantify productivity gains, review-time changes or defect rates, and it does not explain the methodology behind the cost-reduction figure or provide a full technical description of every named tool.
Source: rss
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
