OpenAI describes its forward-deployed engineering team
OpenAI's head of Forward Deployed Engineering says the team embeds with enterprise customers to close the "deployment gap" between model capability and production use, then feeds what it learns back to product and research.
OpenAI has published a post by Colin Jarvis, head of Forward Deployed Engineering (FDE), explaining how the team works. It announces no new product or model. It reads partly as a recruiting pitch, but it includes some specific customer results.
Jarvis says he founded the team in early 2025 to close what OpenAI calls "the deployment gap": the distance between what models appear capable of and what people can reliably use. When models can do a job, FDEs work with Product to make the solution repeatable. When they can't, FDEs work with Research to define the gap. "If we make ourselves obsolete every couple of months, our FDE team is doing its job," he writes.
The most detailed example is an unnamed semiconductor customer. In mid-2025, about 15 FDEs worked on more than 10 use cases aimed at making its chip verification process AI-native. According to the post, verification can account for roughly 70% of a project's effort in that industry. Jarvis says all of the use cases reached production within three to six months. The post reports:
- about 48% time saved through AI-based directed test generation
- a 22% median performance improvement from kernel optimisation
- weeks of debugging compressed into hours
The post also describes John Deere using OpenAI's APIs to give farmers personalized setup and usage guidance for its See & Spray weed-control system. It gives no results for that deployment.
Jarvis adds that as Codex improved, the team moved custom applications onto a common harness built on Codex primitives. The available text ends before he describes the results of that move.