# How 24,000 companies keep their AI from Breaking in Production | Rohit Agarwal, Portkey

Page: https://stenobird.com/podcast/the-neon-show-862453/how-24-000-companies-keep-their-ai-from-breaking-in-production-rohit-agarwal-portkey
Text version: https://stenobird.com/podcast/the-neon-show-862453/how-24-000-companies-keep-their-ai-from-breaking-in-production-rohit-agarwal-portkey.md
Podcast: [The Neon Show](https://stenobird.com/podcast/the-neon-show-862453)
Published: 2026-03-19T12:30:00+00:00
Episode link: https://www.buzzsprout.com/1601695/episodes/18874041-how-24-000-companies-keep-their-ai-from-breaking-in-production-rohit-agarwal-portkey.mp3
Audio file: https://www.buzzsprout.com/1601695/episodes/18874041-how-24-000-companies-keep-their-ai-from-breaking-in-production-rohit-agarwal-portkey.mp3
Processing state: not_requested
JSON: https://stenobird.com/v1/public/podcasts/the-neon-show-862453/episodes/how-24-000-companies-keep-their-ai-from-breaking-in-production-rohit-agarwal-portkey
Duration seconds: 4724

## Resource

Over 1 Trillion AI tokens pass through Portkey every single day. Every AI product eventually runs into the same problem. The prototype works, but once it goes live the system has to manage multiple models, rising token costs, unpredictable latency, and infrastructure that was never built for AI workloads. That is the problem Rohit Agarwal is solving with Portkey, an AI gateway that sits between applications and the models, whether that’s GPT-4, Claude, or Gemini. With 24,000 companies routing...

## Actions

- request_transcript: `POST https://stenobird.com/v1/public/podcasts/the-neon-show-862453/episodes/how-24-000-companies-keep-their-ai-from-breaking-in-production-rohit-agarwal-portkey/transcription-requests` — Idempotently request low-priority transcript generation for this episode.
- read_markdown: `GET https://stenobird.com/podcast/the-neon-show-862453/how-24-000-companies-keep-their-ai-from-breaking-in-production-rohit-agarwal-portkey.md` — Read the agent-friendly Markdown representation of this episode resource.

A page view does not enqueue transcription. Agents should invoke `request_transcript` explicitly when they need this episode processed.

## Transcript

Full transcripts are not published on public pages unless there is a clear rights basis.