What Is a Forward Deployed Engineer (FDE)? The Role OpenAI, Anthropic, and Google Are Fighting Over

In the 2025 job market, one role's posting count grew by an extraordinary 1,165% year over year. That role is the FDE — the Forward Deployed Engineer. The momentum has not stopped in 2026: OpenAI, Anthropic, Google, Databricks, and Scale AI have all stood up dedicated teams, and Google Cloud alone is hiring at a scale of dozens in 2026. Why has a quiet job that Palantir has run for nearly two decades suddenly become "the hottest title of 2026"?

Here is the conclusion up front: an FDE is "an engineer who carries their own company's product into the customer's site and personally owns observation, design, implementation, operation, and product feedback end to end." "Forward Deployed" is a military term meaning "deployed to the front line" — not at the company's HQ R&D division, but stationed at the front line where the customer's work actually happens. Where an ordinary software engineer "receives a spec and builds it," an FDE digs out "problems the customer cannot yet put into words" on site and turns them into product right there.

Let me state my own view first: the FDE is the most misunderstood role of 2026. It is neither "an SE who travels a lot" nor "a consultant who can code." As far as I can see, the essence of the FDE is "a circuit that connects information you can only obtain on the customer's site directly to product decisions." That is exactly why AI companies all want them. This article lays out what an FDE is, why the role exploded in 2026, the actual flow of the work, pay and career, the difference from similar roles, who it fits, and how to get there — all with the latest data. If you are thinking about an AI-engineering career, reading AI API for beginners, the white-collar disappearance debate, and what a multi-agent is alongside it will give you a more three-dimensional picture.

1. What Is an FDE? — The Engineer Who Goes to the Customer's Site

To define the FDE (Forward Deployed Engineer) in one line: "an engineer whose job is to bring their own company's product into the customer's office and land that product in the customer's real, day-to-day work." "Forward Deployed" is a military term for forces stationed in the operational theater rather than at home. The name itself expresses the idea of "separating the engineer from HQ R&D and stationing them at the front line where the customer's work actually happens."

The company that systematized this role nearly two decades ago is Palantir. Its software embeds deeply into the complex data work of government agencies and giant enterprises, and "sell it and walk away" never worked. So Palantir established a model in which engineers are embedded at the customer's site, observing the work and reshaping the product on the spot. This is the prototype of the FDE. At Palantir, the title FDSWE (Forward Deployed Software Engineer) still carries the core of the company.

The difference from an ordinary software engineer is "distance." A typical engineer receives a spec organized by a product manager and implements it. An FDE does not receive a spec. They "discover the spec on site." Inefficiencies the customer cannot yet name, work buried in Excel, legacy integrations nobody wants to touch — the FDE digs these "un-articulated problems" out of the atmosphere of the site and turns them into a prototype on the spot. Running engineering skill and the observational power to understand the customer's work simultaneously inside the same person is the essence of the FDE.

2. Why the FDE Role Exploded in 2026

In the previous section we saw the FDE's face as "an old role that has existed for 20 years." So why has an old role suddenly become "the hottest title" in 2026? The answer lies in the spread of generative AI itself.

Generative-AI products have a structural weakness: "the demo works like magic, yet the moment you bring it into the customer's real work it suddenly stops working." Customer-specific data formats, internal exception rules, integration with existing systems, the workflow of the floor — no matter how smart the AI model is, this "last mile" has to be closed by someone, by human hands. Not sales, not support, but a human who can write code while landing the product in the customer's work. That is the FDE. AI companies struggle to differentiate on raw model intelligence, and the contest has shifted to "can you actually make it run on the customer's site?"

The symbolic move came in late 2025: Anthropic and OpenAI stood up FDE-style teams at almost the same time. In May 2026, Anthropic announced an enterprise structure built around engineers who embed deeply with customers. Google went further — it was reported that in 2026 it would hire FDEs at a scale of hundreds, "stationed inside customer offices and shipping production AI code." Tech-industry trade media call this role "tech's secret weapon" and "the hottest title of 2026."

One caveat. The 2024-2025 FDE boom was partly overheated by "AI FOMO (fear of missing out)." The 2026 market has cooled a notch from there. Today, the companies paying high compensation do so for FDEs who can show, in numbers, how their deployment affected customer retention. The title-only boom is over, and the field is narrowing to "FDEs who can speak in results" — which, I think, is actually a healthy sign for aspiring candidates.

3. The FDE Job — A 5-Stage Loop From Observation to Product Feedback

"An engineer who goes to the site" is hard to picture concretely. One FDE engagement roughly runs as a loop cycling through five stages. What decisively distinguishes it from ordinary contract development is that the final "product feedback" always comes back around.

Picture it concretely. Suppose an FDE joins a large insurance company and takes charge of an AI-powered claims-assessment support tool. In STEP 1, sitting next to an assessor for a week, they realize "the bottleneck wasn't the assessment itself but the time spent searching for similar past cases." In STEP 2-3 they build a similar-case-search prototype in two weeks, and in STEP 4 they make it stick on the floor. Then STEP 5 — they bring it back to HQ saying "a similar-case-search feature is something every insurance company would want," and it gets promoted to a standard feature. Once this lap closes, one FDE's on-site experience moves the whole company's product forward. This is what decisively separates the FDE from "just an embedded engineer."

4. Pay and Career — Why "Future Founders" Flock to It

In the previous section we saw the weight of the FDE's work. The pay duly matches that weight. Palantir FDEs average a TC (total compensation) of about $238,000, with a range roughly from $205,000 to $486,000, and staff-level FDEs reach over $630,000. At OpenAI and Anthropic, packages of $350,000 to $550,000 are becoming standard for mid-to-senior levels. Even by the US-wide median of around $173,816, it is clearly above general software engineering (all figures are US-based, 2026 survey values; rates within other countries are lower).

But what truly attracts many FDE candidates is not the number itself. It is the value of "the best possible training ground for starting your own company next." Within a single engagement, an FDE experiences nearly everything a startup founder does — customer negotiation, problem discovery, design, implementation, operation, and presenting to executives. And they get to peek into the "on-site reality" of multiple industries while being paid. That is why, in recent years, startups have been deliberately hiring "future-founder candidates" as FDEs. In fact, it is not unusual for FDE alumni to go straight into founding companies.

As a career path, three directions open up from the FDE. (1) Product manager or head of product — because no one has richer on-site insight. (2) Solutions architect or engineering manager — because they can speak both technology and customer. (3) Founding a company — as noted above. Those who want to deepen their expertise as an AI engineer can build on knowledge from AI API for beginners and multi-agents, and use the FDE as a few years of training in "being ambidextrous across technology and business."

5. FDE vs Similar Roles — SE, Consultant, Applied AI Engineer

The FDE is often sloppily described as "an SE who is like a consultant," but that misses the essence. Lining it up against four roles that look similar makes the FDE's outline clear.

The pair most easily confused in 2026 is the FDE and the "Applied AI Engineer." The two overlap considerably, but the emphasis differs. An FDE is judged on "depth of deployment" — how far they embedded into the customer's work and made it stick. Palantir and OpenAI favor this title. By contrast, an Applied AI Engineer emphasizes "rigor of AI quality and evaluation" — prompt design, evaluation design, accuracy craft. Anthropic and many AI startups tend to prefer that label. When reading a job posting, looking past the title at "is there embedding at the customer site?" and "are results measured by customer retention?" tells you whether it is a genuine FDE role.

6. Who Fits the FDE Role — and Who Doesn't

The FDE offers both high pay and big growth opportunity, but it is not a role for everyone. The traits we saw earlier — "embedding" and "digging out problems on site" — are selective about people.

Let me say it plainly. People who "want to master one technology deeply" do not need to force themselves toward the FDE. That is not an inferior choice — it is simply a different road. The FDE is a role that delivers value "broadly, fast, on the customer's site," and that points in a different direction from the strength of quietly honing deep expertise. Honestly assessing which one brings you joy is, in the end, the best career judgment.

7. How to Become an FDE — Preparation and What to Learn

So if you actually aim to become an FDE, what should you prepare? What the FDE demands is "the multiplication of three abilities." Not one specialty, but having all three at least at a minimum bar is what counts.

(1) The engineering power to build working things fast. Burst speed to ship a prototype that runs on site within days, rather than beautiful architecture. Touch the full stack broadly, and don't flinch at API integration or data processing. If AI products are the premise, you'll want to have the basics of the AI API and integration tech like MCP down. (2) Problem-discovery and communication power. Draw out what the customer does not say through observation, and speak with the same intensity to executives and to floor staff alike. (3) The power to learn an unknown industry fast. Insurance, manufacturing, logistics — an FDE absorbs the common sense of a new industry in a week per engagement. "Having learned how to learn" is itself a skill.

The realistic first step of preparation is this. First, pick one "un-articulated inefficiency" in your own industry or workplace, build an AI tool to solve it yourself, and carry it all the way to a state where it is actually used. This is a miniature of the FDE's five-stage loop (observe → design → implement → operate → feed back to product). Plenty of people write "implemented the features I was told to" on a résumé, but few can write "found a problem on site myself, turned it into a tool, and it keeps being used." That one line is exactly what works in FDE hiring.

Summary

An FDE — a Forward Deployed Engineer — is an engineer who carries their own company's product into the customer's site and personally owns observation → design → implementation → operation → product feedback end to end. In 2026, OpenAI, Anthropic, and Google all imported this model, which Palantir systematized over roughly 20 years. The reason: generative AI carries a last mile of "the demo works but it doesn't work on site," and the FDE is the role that closes it with human flesh and blood. Postings grew 1,165% year over year in 2025, and pay sits at a high level — Palantir's average of $238K and staff-level over $630K.

The FDE is neither "an SE who travels a lot" nor "a consultant who can code." Unlike an SE, they feed on-site insight back into their own product; unlike a consultant, they write the code through to the end. Where an Applied AI Engineer hones "AI quality," the FDE is judged on "depth of deployment." It fits ambidextrous people who can enjoy ambiguity, and does not fit those who want a clear spec and deep expertise — neither is above the other; they simply point in different directions.

In the end, what the FDE role teaches us is this: "in the age of AI, what keeps its value to the very end is not the model, nor the code, but the human who can stand between the customer's site and the product." A product gains meaning only when it meets the field. The world still does not have nearly enough people who can make that meeting happen. Reading the white-collar disappearance debate, veterans vs juniors, and what a multi-agent is alongside this should let you draw a more three-dimensional picture of an AI-era career.

FAQ

Q. Is the FDE harder than an ordinary software engineer?
A. Less "harder" than "the breadth of abilities required is different." The pure technical difficulty does not differ much from ordinary development, but the FDE stacks customer negotiation, problem discovery, unknown-industry learning, and executive presentations on top. For someone who wants to compete on technology alone the burden is heavy; conversely, for someone who finds "tech alone unsatisfying" it is the finest stage.

Q. Do FDE jobs exist outside the US too?
A. Yes. The numbers are still smaller than in the US, and they center on local arms of foreign AI companies, SaaS startups, and some consulting firms. The title is often "solutions engineer," "customer engineer," or "implementation engineer," so it helps to read postings by the reality of "embedded at the customer site, writes code themselves, and is also involved in product improvement." Pay levels are lower than the US figures (such as $238K) and follow local engineering rates.

Q. Can you become an FDE with no experience, straight out of school?
A. Few companies have new-graduate FDE tracks, so it is realistic after 2 to 4 years of software development experience. New graduates should first build full-stack implementation skill in a regular engineering job, and in the meantime create one track record of "found a problem on site myself and turned it into a tool." Aiming for an FDE role from there is the solid route.

Q. Won't the FDE be replaced by AI?
A. The FDE is on the side that uses AI to the hilt. Prototype implementation gets many times faster with AI. But the parts of "digging out problems the customer cannot articulate from the atmosphere of the site" and "earning the trust of executives and the floor at the same time" are hard for AI to replace. In fact, the faster AI makes implementation, the more the bottleneck shifts to "the power to find the right problem," and the FDE's value trends upward.

Q. What is the single biggest difference between an FDE and an SI (systems integrator)?
A. "To whom they return the learning." An SI builds a system for the customer and the engagement ends. An FDE builds for the customer and, at the same time, returns the insight gained there to their own core product. One FDE's on-site experience moves the company's product roadmap — whether this "lap back to the product" exists is the decisive difference.