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 details how the company’s existing developer infrastructure and newer AI tools support engineers using coding agents. Uber says it uses a model gateway, shared tools for multiple coding harnesses, spending controls and human review; productivity effects and several operational details are not quantified in the supplied report.
Uber’s AI infrastructure combines internal developer systems with newer tools for coding agents, according to a report based on an interview with Uday Kiran Medisetty, a Distinguished Engineer at the company. The account describes how Uber connects agents to its code, services and engineering knowledge while retaining controls for model choice, spending and engineer-led code review.
The interview focuses on tools engineers use in their daily work, including DevPods, Code Inbox, Slate and Wisdom, among more than 20 internal tools mentioned in the report. Medisetty discusses background agents, tools for development and bug fixes, evolving code review, token costs and how Uber measures productivity as engineering work changes. The supplied report excerpt does not explain the function of each named tool in detail.
Uber’s platform team decided in 2023 that AI systems should be model- and harness-agnostic, the report says. Its Model Gateway supports models from Anthropic and OpenAI, as well as open models Uber runs. The company’s AI Guard middleware checks requests and responses, while the gateway also supports spending records, audit logs and session traces. Uber’s AIFX wrapper supports Claude Code, Codex and OpenCode, carrying settings, skills and plugins across those harnesses.
Uber assigns engineers AI spending tiers, according to the report. A manager can approve additional spending when an engineer reaches a limit; a CostCheck plugin offers suggestions for managing costs across sessions. The report says Uber has reduced token costs by about 50% in recent months, but the excerpt does not give a precise measurement window or explain the calculation. It also says all pull requests continue to receive human review, with AI review, relevant screenshots or videos, and risk scores intended to help engineers assess changes.
How Uber Keeps Agents Governed
The account shows that deploying coding agents at a large company involves more than choosing a model. Uber’s Model Gateway is intended to let teams change providers without tying their infrastructure to one vendor, while shared harness settings and spending controls offer ways to manage how engineers use the tools.
That matters because agents interact with company code and services, and their use can carry costs and review demands. Uber’s approach, as described in the interview, combines automation with security checks and human approval rather than treating generated code as ready to merge. The report also identifies code review as a bottleneck, though it provides no measured evidence about whether the tools have reduced review time or improved output.
AI coding assistant tools for developers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Infrastructure Built Before AI Agents
The report places the AI tools within Uber’s longer history of building developer infrastructure for its scale. Medisetty joined in 2015, when the company had two services, according to the account. By 2019 it had thousands; the report says the service count in 2026 is expected to be about the same as in 2019, despite the business being roughly 10 times larger. Those figures are reported estimates, not independently verified in the supplied material.
Uber developed internal systems when existing products could not meet its needs, the report says. Examples include uOwn for service ownership, M3DB for high-volume metrics, Cadence for durable workflows and Morpheus for feature flags and experimentation. The account also describes cities as a basic operating unit: city-level requirements affect observability, deployments, feature flags and infrastructure, with cities assigned to cloud regions that can fail over to another region.
That earlier investment helps explain why the company’s agent tooling is presented as part of a broad internal platform rather than a standalone chatbot. The interview links agents to existing code and engineering knowledge, but the supplied excerpt does not describe the architecture or access permissions of those connections in technical detail.
“The report says Uber’s platform team decided in 2023 that everything built with AI should be model- and harness-agnostic.”
— Uday Kiran Medisetty, Uber Distinguished Engineer, as described in The Pragmatic Engineer interview
internal developer infrastructure software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What the Interview Does Not Measure
The supplied report does not provide a detailed accounting of the tools’ adoption, the number of engineers using agents, or the measured effect on delivery speed, defect rates or review workload. It says Uber is working on how to measure productivity as the engineer’s role changes, but does not give a methodology or results.
The reported 50% token-cost reduction lacks a stated comparison period and baseline in the excerpt, so it cannot be translated into a specific savings figure. Details about the risk scores, AI review system, agent permissions and the individual functions of several internal tools are also not included. The claims and descriptions are presented through an interview and report, not as independently audited performance results.
As an affiliate, we earn on qualifying purchases.
How Uber Will Track Agent Use
The report points to ongoing work on measuring productivity as AI changes engineering tasks, but it does not name a release date, target or next public milestone. Further evidence would be needed to establish whether the tools measurably change code-review times, engineering output, reliability or total AI spending.
For now, the next developments to watch are how Uber expands its supported models and harnesses, how its internal spending and security controls evolve, and whether the company publishes metrics that clarify the effects of agent use. The supplied material does not state when any such update will be released.
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 developer tools and systems that support AI coding agents.
Which AI models and coding tools does Uber support?
The report says Uber’s Model Gateway supports Anthropic, OpenAI and open models run by Uber. Its AIFX wrapper supports Claude Code, Codex and OpenCode, carrying settings and plugins between them.
Does AI review replace human review at Uber?
No. The report says engineers still review every pull request. AI review and risk scores are among the tools intended to help reviewers evaluate changes.
How does Uber manage AI coding costs?
According to the report, engineers receive spending tiers, managers can approve additional spending, and the CostCheck plugin offers cost-management suggestions. The report also claims token costs fell about 50% in recent months, without giving a precise measurement window in the supplied excerpt.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
