OpenAI Forward Deployed Engineer: the job, the pay and the interview, from OpenAI's head of FDE (2026)

An OpenAI Forward Deployed Engineer (FDE) embeds with one of OpenAI's largest customers, takes a hard problem from scoping to production, and feeds what breaks back to OpenAI's Product and Research teams. In October 2026 the head of the team, Colin Jarvis, published what the job means in his own words (What FDE means at OpenAI). His shortest version: "If we make ourselves obsolete every couple of months, our FDE team is doing its job." Below is what the post says, what OpenAI's live postings add, and what both tell you about the interview.

The job in one paragraph

Jarvis founded the team in early 2025 to close what he calls "the deployment gap": the distance between what models appear capable of and what people can reliably use to do their work. Where the model can already do the job, FDEs work with Product to make the solution repeatable. Where it cannot, they work with the customer and Research to define the gap and close it. The stated goal of every engagement is that the next customer starts from a better floor.

The postings say the same thing in job-description language. The San Francisco FDE posting asks you to "own discovery, technical scoping, system design, build, and production rollout" and to "measure success through production adoption, measurable workflow impact, and eval-driven feedback that changes product and model roadmaps."

Four things the post shows about the work

1. Start with the outcome. Jarvis opens every team session with one slide: "Start with the outcome, do what it takes to succeed, then follow with scale." His point is that a demo can go a long way before anyone asks whether it works. The bar is a standard the team "can stand behind in production."

2. Learn the industry on the job. In mid-2025 about 15 FDEs worked with a large semiconductor customer on 10+ use cases (directed test generation, root cause analysis, an architecture assistant). Some had joined OpenAI days earlier and had never worked in semiconductors. They studied chip design in their spare time and built evals with the customer's engineers. Within 3 to 6 months every use case reached production, with about 48% time saved on directed test generation. His line: "Our FDEs don't need to know the industry on day one but it is their responsibility to learn it."

3. Delete your own code when the product catches up. When Codex improved, one FDE rebuilt the applications around it and the team moved the use cases onto a common Codex harness with skills, plugins and MCP connectors. About 90% of the custom code per project went away. In one August 2026 case the first deployment took two months and the second team took one day.

4. Say when the model cannot do it yet. A telecom customer wanted to extend customer service to voice, which was about 90% of its volume. During scoping the model scored about 60% on the key evals. The team told the customer, brought Research in, and tested successive model snapshots against the customer's requirements. Over six months the evals went from about 60% to 80% and then above 90%, enough for production.

Who OpenAI is hiring right now

Read from OpenAI's own Ashby board on 2026-10-26: 71 open roles with Forward Deployed, FDE or Applied AI Engineer in the title (30 forward deployed titles, 41 applied AI), across 15 cities. 31 are in the US (San Francisco, New York, Seattle, Washington DC) and 40 are outside it (London, Paris, Munich, Zurich, Madrid, Tokyo, Seoul, Singapore, Sydney, Sao Paulo, Abu Dhabi).

Pay is printed on all 31 US roles and on none of the 40 outside it. The US bands:

• Forward Deployed Engineer (general, Healthcare, Legal, Financial Services): $185K to $300K plus equity.
• Forward Deployed Software Engineer: $153K to $325K (New York, Seattle), $185K to $325K (San Francisco), $145.8K to $280K (Government, DC).
• Deployment Lead, FDE: $230K to $294K. Manager, FDE (New York): $302K to $385K.
• Forward Deployed Security Engineer (DC): $266K to $445K, the top of the range.

The SF FDE posting asks for 5+ years of engineering or technical deployment experience that includes customer-facing work, production code in Python or JavaScript, and experience building systems on LLMs. It is hybrid, 3 days a week in office, with travel up to 50%.

One distinction that trips people up: OpenAI's FDE team is not the same organization as The OpenAI Deployment Company (DeployCo). OpenAI's FDE page says its FDEs connect customer engagements with Product and Research, while DeployCo "brings additional engineering and deployment capacity to support implementation at scale." DeployCo hires on its own board with its own process; our free DeployCo loop dossier covers it.

What this means for the interview

OpenAI's public interview guide describes the shape: an intro call, a skills assessment (pair coding, a take-home project or a technical test, sometimes more than one), then 4 to 6 hours of final interviews with 4 to 6 people over one or two days. The guide's own summary of what they look for: "We want to understand how you think." The post tells you what that thinking should look like. Prepare four things:

1. An outcome before a design. When you get a vague customer problem, name the business result and how you would measure it before you draw any architecture. That is the slide.
2. The eval. Every story in the post turns on an eval: chip-test quality, the voice evals at 60% and then above 90%. Be ready to say what you would measure, on what data, and what number makes it production-ready.
3. The honest gap. Have one story where the model or the tool could not do the job and you said so, then changed the plan. The telecom story is the template.
4. Learning a domain fast. They hire people who were new to semiconductors and shipped in months. Bring a time you learned a customer's field quickly and what you built because of it.

The round that tests all four at once is the customer case: a vague problem, a clock, and an interviewer playing the customer. You can practise one now, free and with no signup, in the Case-Study Arena. It grades how you break the problem down against a hidden rubric.

OpenAI's roles sit beside every other company hiring the same job on our live board, each checked against the employer's own careers system, with the published pay and a direct apply link.