The AI Record
And we have this notion of spend tiers. So by default, you get certain … tier, and then we have many levels of tiers. Once you're close to that tier, then you can raise a request to go to the next tier. And you just need a simple manager approval. … But we don't want to restrict engineers … they should be able to spend whatever they need because ultimately we want to … maximize the gain from this.

Uday Kiran MedisettyDistinguished Engineer, Uber · Practitioner, Uber

Context. Answering the host's question whether an engineer at Uber needs to be aware of AI spend.

Podcast: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 26:54. Reported by Gergely Orosz.

Checked against the source [ Grade C ] Said in the last year Uber: $52 billion revenue, FY2025

Filed under

Who is speaking

Uday Kiran Medisetty, Distinguished Engineer. Runs the engineering-productivity work Medisetty reports on and gave the AI Engineer World's Fair talk; an insider with an interest in presenting the program well.

How this was checked

Checked against the source by a first reader, then confirmed independently by a second reader who fetched the source without seeing the first reader's notes (Independent review agent (Claude Sonnet), blind second pass from passage text and citation only, no repository access, 2026-11-07).

Whisper transcript, around 27:00. 'certain, certain tier' is covered by the ellipsis, as are the skipped sentences after 'manager approval' and 'restrict engineers'. One quirk: the passage has an ellipsis in 'we want to … maximize the gain', but the source has no omitted words there ('we want to maximize the gain'). No wording is altered.

How to cite this passage

Uday Kiran Medisetty (Uber), “Uber's AI infra, from the inside”, The Pragmatic Engineer (YouTube), 2026-10-14, at 26:54. https://www.youtube.com/watch?v=MDUYSMJmi8A Archived at https://the-ai-record.com/passages/uber-spend-tiers-manager-approval-roi-conversation-2026

More from the same source

what we call minion internally. So this is … background coding agent. So around 15% of all of the production … go through minion right now. … Once you have for the task that I have really good spec, or I want to do minimal steering, I can hand off that to a minion.

Uday Kiran MedisettyDistinguished Engineer, Uber · Practitioner, Uber

Context. Medisetty is describing Uber's background coding agent, Minion. The word cut after 'production' is pull requests (PRs), which both transcriptions rendered as 'peers'; he says 'PRs' in the next sentence.

Podcast: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 50:46. Reported by Gergely Orosz.

  • 15% production pr share via minion, as of the October 2026 recording. Level not stated

Checked against the source [ Grade C ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

year over year, the number of CI jobs also doubled or more than doubled. So we had to like also like beef up whole bunch of foundations in this space

Uday Kiran MedisettyDistinguished Engineer, Uber · Practitioner, Uber

Context. Speaking about readying Uber's systems for the influx of pull requests that agents produce.

Podcast: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 58:38. Reported by Gergely Orosz.

  • 2 x ci jobs growth, year over year, as described in the October 2026 recording

Checked against the source [ Grade C ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

And since all the requests go through one place, we could also add like audit logging, session traces, and a whole bunch of infrastructure around this. Every model request at Uber goes through this. So there's no model request that doesn't go through this. So that gives us like one single entry point for across all the models.

Uday Kiran MedisettyDistinguished Engineer, Uber · Practitioner, Uber

Context. 'This' is Uber's internal model gateway, which he has just been describing (one OpenAI-compatible or Anthropic-compatible API in front of models from all vendors, with an AI Guard safety layer).

Podcast: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 24:22. Reported by Gergely Orosz.

Checked against the source [ Grade C ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

So like one, for example, one of the things we did is tagging all of our PRs to different categories, like on complexity level, and is it a feature PR bug fix maintenance testing documentation on different categories. And we could see year over year, how is that mix changing? Luckily, like at least like right now, you can see our percentage of feature PRs is going up, our percentage of maintenance PRs is going down.

Uday Kiran MedisettyDistinguished Engineer, Uber · Practitioner, Uber

Context. Answering the host's question how Uber thinks about engineering productivity now that agents write more code. He gives no figures for the shift.

Podcast: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 1:24:33. Reported by Gergely Orosz.

Checked against the source [ Grade C ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

So it's been really useful, even in our support channels, I would see like someone reports a bug, the on call ask minion, hey, like, … what is the issue and then minion will come back with, oh, this is the root cause, oh, this looks simple enough, let's just fix it right there and then we would like fix it, like there's no, then loop of like adding it to backlog, doing in the next, next thing, like you're able to fix it right there.

Uday Kiran MedisettyDistinguished Engineer, Uber · Practitioner, Uber

Context. Medisetty is describing Slack use of Minion, Uber's background coding agent.

Podcast: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 53:00. Reported by Gergely Orosz.

Checked against the source [ Grade C ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

right now we still require an engineer to review every PR … goes in. … Once the more confidence we get in our risk scoring and we can also back test it with years of data, right? We have all of this data and the more confidence we have in our AI code reviews, then maybe for some class of low risk PRs, we could enable, we call it green lane like auto landing.

Uday Kiran MedisettyDistinguished Engineer, Uber · Practitioner, Uber

Context. Answering the host's question whether human code review will always exist.

Podcast: Uber's AI infra, from the inside, The Pragmatic Engineer (YouTube), October 14, 2026, at 59:37. Reported by Gergely Orosz.

Checked against the source [ Grade C ] Said in the last year Uber: $52 billion revenue, FY2025 Passage page

Something wrong? Report a correction. JSON: /api/v1/passages/uber-spend-tiers-manager-approval-roi-conversation-2026.json