What is a forward deployed engineer (FDE)? Meaning and role
FDE meaning: a forward deployed engineer is a software engineer who works at the customer. Origins at Palantir, role, skills, salaries, versus consulting roles.
A forward deployed engineer (FDE) is a software engineer who works at the customer's site, inside the customer's teams, to build a solution on the customer's data and systems and take it to production. French has no settled translation: Anthropic's job posting in Paris calls the role « ingénieur en déploiement avancé », roughly "advanced deployment engineer".
The role comes from Palantir, which claims to have pioneered it. Generative AI changed its scale: according to the Financial Times, FDE job postings rose by more than 800% between January and September 2025, and on 6 November 2026, OpenAI listed 32 FDE openings, including one in Paris.
An FDE writes the code that goes into production at the customer: that is what separates the role from consulting and pre-sales engineering. For a CIO or an executive, the practical question is when this way of working pays off, and who to do it with.
What does FDE mean?
FDE is short for forward deployed engineer. The engineer is "deployed forward", sent as close to the ground as possible: into the customer's offices, plants or tools, rather than the employer's product team. The FDE learns the need on site, builds the solution with the users and carries it through to production.
Gergely Orosz, author of The Pragmatic Engineer newsletter, gives a tighter definition: a software engineer who alternates between being embedded with customer teams and the core product engineering team. That back-and-forth matters: what the FDE learns at one customer flows back into the product. OpenAI's Paris posting asks for exactly that: sharing field feedback with Research and Product.
Two vocabulary notes:
- FDSE (forward deployed software engineer) is Palantir's historical title; on 6 November 2026 Palantir listed 44 Forward Deployed Software Engineer openings. OpenAI uses both forms; Anthropic and most others say FDE. It is the same job.
- In IT security, FDE also stands for full disk encryption, which has nothing to do with the role.
Why "forward deployed"? The military origin of the term
The phrase comes from US military jargon, as the New York Times points out. A forward-deployed force is stationed far from its home bases, close to where it may have to act, so it can respond without delay. NATO applies the same principle with its Forward Land Forces: nine multinational battlegroups stationed on its eastern flank, from Finland to Bulgaria.
The parallel with the job is direct. The FDE leaves the employer's headquarters to settle in at the customer, the way a forward unit leaves its base: they see the problem up close and decide on the spot, without waiting for a product team to rule from a distance. The term also fits Palantir's customers: the company still advertises Forward Deployed Software Engineer "Tactical Edge" roles open only to holders of a US Top Secret security clearance.
How do you say forward deployed engineer in French?
There is no established translation. « Ingénieur déployé chez le client » (engineer deployed at the customer) describes the role best; « ingénieur embarqué » (embedded engineer) is also used. Anthropic went with « ingénieur en déploiement avancé », but Palantir keeps « Forward Deployed Software Engineers (FDSE) » in its French-language internship posting, and Mistral AI advertises under the English title. In France, the English term and the FDE acronym remain the norm, job ads included.
Where does the forward deployed engineer role come from?
The role was born at Palantir, the US data analytics software company founded in 2003. According to The Pragmatic Engineer, it was created in the early 2010s under the internal name "Delta", and until around 2016 Palantir had more FDEs than regular software engineers. The Hudson Institute credits Shyam Sankar, now Palantir's chief technology officer, who envisaged the role of the forward deployed engineer after joining as employee number 13.
The in-house jargon talks of "Deltas" and "Echoes", according to a16z. The Delta is the engineer who codes. Alongside, Palantir hires Deployment Strategists, described in the job posting as people who immerse themselves in customers' most intricate workflows: business and engineering, side by side at the customer.
The model is still central at Palantir: on 6 November 2026, 73 of its 309 open positions had "Forward Deployed" in the title, by our count.
Why generative AI made the role explode
A language model is a generic component. To be useful inside a company, it has to be connected to the right data, respect access rights, have its answer quality measured, fit into a process and win over the teams who will use it. Colin Jarvis, who leads OpenAI's forward deployed engineers, says deployment is the problem in about 80% of cases, not model capability. The bottleneck has become integration, which is precisely an FDE's job.
So the model makers hired. At OpenAI, Colin Jarvis pitched an FDE team at the beginning of 2025; it started with two engineers and had more than ten by August 2025. On 6 November 2026, OpenAI listed 32 FDE openings out of 823 open roles, in fourteen cities including San Francisco, Paris, Munich, Zurich and Tokyo, with healthcare, legal, financial services and cybersecurity specialisms. Anthropic listed eight, two of them in Paris. Mistral AI is hiring an "Applied AI, Forward Deployed Machine Learning Engineer" in Paris, and French startups such as H Company and Linkup have opened FDE roles in Paris.
Market data points the same way. Besides the 800% rise reported by the Financial Times, The Pragmatic Engineer relays a Wall Street Journal count of more than ten times as many Indeed postings in 2025 as in 2024. As early as June 2025, a16z called it the hottest job in startups, arguing that AI agents only deliver value with real implementation work at the customer.
What does a forward deployed engineer do day to day?
Job postings describe very concrete work. OpenAI's Paris posting asks the FDE to own technical delivery from first prototype to stable production, embed closely with customer teams, make trade-offs between scope, speed and quality, then codify what works into reusable tools. Anthropic's Paris posting names specific deliverables: MCP servers, sub-agents and agent skills used in production workflows, and the identification of repeatable deployment patterns. MCP, the protocol that connects a model to a company's tools, is covered in our article on the Model Context Protocol.
FDEs spend a large share of their time at customers: up to 50% travel at OpenAI, 25 to 50% at Anthropic in Paris, up to 25% at Palantir.
What skills does an FDE need?
Postings converge on five skills:
- Writing production code, mostly in Python and TypeScript, not just prototypes.
- Running LLMs in production: prompting, evaluation, RAG, agents. Mistral AI asks for experience with fine-tuning and advanced RAG or agentic use cases.
- Learning a business fast enough to talk with a financial controller, a lawyer or a plant manager.
- Working with ambiguity: the need is never written down at the start.
- Communicating, often in two languages: for Paris, Anthropic requires native French and fluent English.
Required experience varies a lot: one year at Palantir, two at Mistral AI, five at OpenAI in Paris, eight at Anthropic in Paris.
FDE vs consultant, solutions engineer and developer: what's the difference?
The FDE borrows from several jobs. According to The Pragmatic Engineer, consultants make one-off recommendations while FDEs work with customers over time and contribute to the product; at OpenAI, solutions architects advise and rarely write code on customer infrastructure. At Anthropic, FDEs work with the post-sales teams, so after the contract is signed. At Palantir, "Devs" build the product and "Deltas" deploy it, as the title of a post on the company's blog sums up.
The table below is our reading of these roles:
| Role | When they step in | What they deliver | Writes production code? |
|---|---|---|---|
| Forward deployed engineer | After the sale, over time | A solution in production at the customer | Yes |
| Consultant | At scoping, one-off | Recommendations, a roadmap | Rarely |
| Sales engineer / solutions engineer | Before the sale | A demo, a proof of concept | Little |
| Solutions architect | Before and during the project | An architecture, advice | Rarely |
| Product developer | Continuously, at the employer | Features of the standard product | Yes, outside the customer's context |
An FDE is judged on what runs in production; a consultant on the report, a sales engineer on the demo. Colin Jarvis says as much for OpenAI: his FDEs are measured on whether projects reach production and move a real metric.
How much does a forward deployed engineer earn?
The only reliably published salaries come from job postings at the large vendors, mostly American:
| Employer | Role | Location | Advertised annual salary |
|---|---|---|---|
| Palantir | Forward Deployed Software Engineer | New York | $135,000 to $200,000 |
| OpenAI | Forward Deployed Engineer | San Francisco | $185,000 to $300,000, plus equity |
| Anthropic | Staff Forward Deployed Engineer | New York, San Francisco, Seattle | $280,000 to $320,000 |
| Anthropic | Forward Deployed Engineer | London | £225,000 to £255,000 |
| Anthropic | Manager, Forward Deployed Engineering | Paris | €235,000 to €295,000 |
These are base salaries collected on 6 November 2026, excluding bonuses and equity unless stated. In France, OpenAI's and Anthropic's Paris FDE postings show no salary, and neither do those of Mistral AI, H Company, Linkup or Ask This Guy. The US labs' ranges say nothing about the French market: they pay senior profiles that these labs recruit in a dozen countries at once.
How to become a forward deployed engineer: training and career path
None of the postings we read asks for a degree specific to the role. Palantir hires people with a background in computer science, mathematics, physics or data science and at least one year of experience; Anthropic asks for a bachelor's degree or equivalent experience. The typical path starts in software engineering, with customer-facing experience: Anthropic mentions consulting and encourages former technical founders to apply.
There are entry points for graduates: Palantir offers an FDSE internship in Paris, Mistral AI an internship in its Applied AI team. For working FDEs, a16z runs an eight-week cohort, mostly a networking program.
For a company, the training question is different: the teams who will take over the solution must upskill during the project, not after the FDE leaves. That is the purpose of AI training built on your own use cases.
How can a French company work with FDEs?
A French company has three options.
- The model vendor's FDEs. OpenAI, Anthropic and Mistral AI are hiring in Paris to support their large customers. Their model expertise is unmatched, but their FDEs deploy their employer's models, and Colin Jarvis admits it: "we should always be temporary". For a company that cares about digital sovereignty, the choice of model and hosting should not depend on who installs it.
- Hiring its own FDEs. The most durable route, but the talent pool is thin: a recruiter quoted by The Pragmatic Engineer says only about 10% of the market wants the role.
- Hiring a model-independent firm, which works inside your teams with the models that fit your context and hands over the code.
The third route is the one we practise at Ask This Guy: our engineers analyse the need with your teams, write the code on your data and stay until the tool is actually used. The approach, engagement models and scope are on our forward deployed engineer page.
Whatever the route, the FDE model pays off when four conditions are met: a problem that matters to the business, accessible data, a sponsor who makes decisions and a measurable success criterion set before development. It does not pay off for a task that fixed rules can handle: there, using as little AI as possible remains the best answer. And if a project is already stuck after the demo, the most common causes are covered in our article on RAG mistakes after the POC.
The limits of the FDE model
The model has critics, and they make serious points. Marc Andrusko, a partner at a16z, warns against thousands of bespoke deployments that are impossible to maintain or upgrade when no real product sits underneath, and against companies relabelling pre-sales or junior staff as FDEs. For lower-stakes problems, he writes, the return does not justify months of on-site engineering.
On the candidate side, enthusiasm does not always match the job-posting numbers. Gergely Orosz describes a role that is more like solutions engineering than software development, and quotes a former FDE put off by "a typical IT services mindset".
For the customer, two risks need managing from day one: dependency, if the knowledge stays in the FDE's head, and drift into an endless engagement. A good FDE plans the exit on arrival: documentation, code handover, team training.
Frequently asked questions about the forward deployed engineer role
What does FDE mean in IT?
In AI and enterprise software, FDE means forward deployed engineer: an engineer who works at the customer to deploy and adapt a solution all the way to production. In IT security, the same acronym also stands for full disk encryption, according to the NIST glossary.
What is the difference between an FDE and an FDSE?
None in substance. FDSE, for forward deployed software engineer, is the historical title at Palantir, which created the role. OpenAI uses both forms, while Anthropic and most other companies say FDE.
How much does a forward deployed engineer earn in France?
There is no reliable public data for the French market: the startups hiring FDEs in Paris, such as Mistral AI, H Company and Linkup, do not show salaries in their postings. The only ranges published in Paris come from US labs, such as Anthropic, which advertises €235,000 to €295,000 for a manager role. In the US, postings range from $135,000 at Palantir to $320,000 at Anthropic.
Do you need a specific degree to become an FDE?
No. Postings ask for an engineering or science background (computer science, mathematics, physics, data science) or equivalent experience, then one to eight years of experience depending on the employer. Customer-facing work and shipping LLMs to production matter more than a specific degree.
Which companies hire FDEs in France?
In November 2026, OpenAI and Anthropic list FDE roles in Paris, as does Mistral AI under the title "Applied AI, Forward Deployed Machine Learning Engineer", along with startups such as H Company and Linkup. Palantir and Mistral AI offer internships in Paris.
How much time does an FDE spend at the customer?
Postings usually mention 25 to 50% travel to customer sites: up to 25% at Palantir, 25 to 50% at Anthropic in Paris, up to 50% at OpenAI. The rest of the time is spent with the employer's product team or working remotely.
Forward deployed engineer is a new title for an old idea: put the person who writes the code next to the person who has the problem. Generative AI made it urgent, because a model is useless until it is wired into a company's data, permissions and habits. For any organisation, the title matters less than three questions: who writes the code, who stays when the contractor leaves and who chooses the model.
At Ask This Guy, we deploy FDEs at our own customers too. To us it is a condition for an AI project to succeed: the vendor has to know its customer and their business, not just the technology. Our work sits where technology meets the business, and our forward deployed engineer offer describes how we organise it.
