A team at OpenAI embeds itself inside some of the world’s biggest companies to turn AI models into real-world deployments.
Colin Jarvis, who leads OpenAI’s forward-deployed engineering team, explained in an episode of the “Altimeter Capital” podcast published Thursday how his team helps companies generate “tens of millions to sometimes the low billions” in value.
The team is still small: 39 engineers, with plans to grow to 52 by year-end, Jarvis said. OpenAI lists 24 openings for the forward-deployed engineering team in the US, Europe, and Japan, with salaries in the US topping out at $345,000, plus equity, according to the job postings.
The term “forward-deployed engineer” was popularized by Palantir, the defense tech software giant. It describes engineers who work directly with clients to fine-tune the product on-site.
When ChatGPT came out in 2022, the model sparked “tons of hype.” “People were really excited, but it was also, like, kind of hard to get value from the models,” Jarvis said.
Early enterprise customers struggled to translate that excitement into usable systems. Jarvis said the only consistently successful approach was to embed directly with clients, learn their workflows, and work alongside their staff. This led OpenAI to set up a forward-deployed model.
One of the team’s major projects was with Morgan Stanley, which became one of OpenAI’s first enterprise customers to deploy GPT-4.
The technical scaffolding took six to eight weeks, but convincing financial advisors to trust the tech took far longer, Jarvis said. The team had to spend another four months running pilots, collecting evaluations, and iterating with wealth advisors.
“In the end, about 98% of them adopted it,” he said.
The team also worked with a semiconductor company in Europe to build a “debug investigation and triage agent” that could examine failures and fix bugs. They looked across the company’s value chain and realized engineers were spending 70% to 80% of their time debugging chips, Jarvis said.
Jarvis said that forward-deployed engineering teams have to be clear about their purpose. His team avoids “services revenue” and is focused on creating product playbooks, he added.
Forward-deployed engineering model
Earlier this year, Jarvis announced in a LinkedIn post that he would be leading OpenAI’s new forward-deployed engineering function.
“Our focus is getting our customers to production, whether it’s through a zero-to-one novel application of our tech or helping you to scale proven cases,” he wrote in January.
Since then, OpenAI has been hiring forward-deployed engineers around the world, including San Francisco, New York, Dublin, London, Paris, Munich, and Singapore.
In July, OpenAI’s international managing director, Oliver Jay, said that the forward-deployed engineering model is a “really specific way to advance the acceleration of advanced AI into scale production cases.”
“This is where we solve the latest gap between companies,” Jay said in Singapore at the Fortune Brainstorm AI 2025 conference.
Venture investors have also noticed the value the model delivers.
YC partner Diana Hu said in an episode of the “Y Combinator” podcast published in June that she and her team have seen founders close “six, seven-figure deals” with major companies by being forward-deployed engineers.
YC CEO Garry Tan also said on the podcast that the model gives AI startups an edge, helping them beat out giants like Salesforce, Oracle, and Booz Allen.
