From the source
OpenAI's head of Forward Deployed Engineering describes how the team works alongside customers to close the gap between what models appear capable of and what people can reliably use.
- Colin Jarvis founded the team in early 2025 to address what he calls "the deployment gap," and says the team's job is to make itself obsolete as products absorb the work.
- For John Deere, OpenAI's APIs were deployed to give farmers personalized setup guidance and recommendations for its See & Spray precision weed-control technology, which uses cameras and machine learning to cut chemical use by up to 70%.
- With a large semiconductor customer, around 15 FDEs worked on more than 10 use cases in mid-2025, embedding with the verification team and building evals; all reached production within the first 3-6 months.
- The reported payoff included roughly 48% time saved through AI-based directed test generation, a 22% median performance improvement through kernel optimisation, and debugging compressed from weeks to hours.
- Jarvis says FDEs need not know the industry on day one but must learn it, and that teams mix generalists with domain experts, as in life sciences where scientific backgrounds help with pre-clinical drug discovery tasks.
