What is a Forward Deployed Engineer?

FDE stands for Forward Deployed Engineer — a software engineer who embeds with a customer and builds the custom work that makes their employer’s platform solve that customer’s problem. It is a real engineering job, not a pre-sales one, and it has become one of the fastest-growing titles in AI.

The confusion around this role is understandable. The title is barely fifteen years old, it means slightly different things at Palantir than at OpenAI, and it sits beside four other titles — Solutions Architect, Sales Engineer, Applied AI Engineer, Implementation Engineer — that overlap with it and with each other. This page is the plain-English version: what the role is, where it came from, what the work is actually like, what it pays, and how to tell it apart from its neighbours.

What FDE stands for

FDE = Forward Deployed Engineer. You will also see it written as Forward Deployed Software Engineer, abbreviated FDSE — Palantir uses both, with FDSE indicating the more code-heavy of the two tracks. “Forward deployed” is borrowed from military logistics, where it describes personnel stationed at the point of operations rather than at headquarters. The metaphor is the whole idea: the engineer goes to where the problem is.

Because the title is not standardised, the same job is advertised under a lot of other names — Deployment Engineer, Delivery Engineer, Deployment Strategist, Field Engineer, Customer Engineer, or simply “Software Engineer” with a customer-facing description buried in the third paragraph. In the postings we classify, about 8% of roles that are substantively FDE work don’t carry “FDE” or “forward deployed” anywhere in the title.

Where the role came from

Palantir originated the term and the staffing model behind it. The company’s software was capable but genuinely difficult to adopt: every customer had different data, in different systems, under different rules, and no amount of documentation closed that gap. Rather than treat this as a services problem to outsource, Palantir treated it as an engineering problem and sent engineers into the customer’s building.

The unit of work became the pod — a small cross-functional team, often a handful of engineers plus a delivery lead, assigned to one account for the length of a deployment and accountable for whether that account actually succeeds. Other companies run the same structure under other names: deployment team, delivery squad, customer engineering team. What makes it a pod rather than a project is that the same people own the outcome from first workshop to production, instead of handing it down a chain.

The model spread because the conditions that produced it spread. Any company selling something powerful, general-purpose and hard to adopt eventually discovers that the bottleneck is not the product but the distance between the product and a customer’s messy reality — and that closing that distance takes engineers.

What a Forward Deployed Engineer actually does

A realistic week has four kinds of work in it, in roughly this proportion:

• Building. Custom integrations, data pipelines, internal tools, and application logic on top of the core platform. This is genuine production code with genuine consequences, and it is the largest share of the job.
• Diagnosing. Working out what the customer actually needs, which is rarely what they asked for. Problems arrive as complaints — “the numbers are wrong”, “this is too slow” — and turning those into a specification is the engineer’s job, not the customer’s.
• Integrating. The hard part is almost never the algorithm. It is the customer’s decade-old identity system, their data residency obligations, their change freeze in December, and the three teams who must agree before anything ships.
• Explaining. Making a trade-off legible to people who control the budget. An FDE who cannot do this can build the right thing and still watch it go unused.

What separates it from a platform engineering role is the feedback loop. You see the person your software fails, usually within a day, sometimes across a table. What separates it from a sales engineering role is that you own delivery: the demo is not the deliverable, the working system is.

The FDE model, and “FDE as a service”

From the employer’s side, the forward deployed model is a deliberate trade. Embedding engineers with customers is expensive and does not scale linearly — but it produces two things that marketing cannot buy: deployments that survive contact with reality, and a direct, unfiltered account of why the product is hard to use. At companies that run this model well, the FDE organisation is the most reliable source of product roadmap in the building, because it is the only function that is measured on whether the thing actually worked.

The usual sequence — sometimes called the FDE process — runs from a scoping workshop, to a narrow first use case chosen because it can ship in weeks, to a working prototype in the customer’s real data, to hardening and handover. The discipline is in the second step: forward deployed teams win by getting something real in front of users early, then expanding, rather than by specifying the whole programme up front.

“FDE as a service” describes the commercial version of this: a vendor supplying forward deployed engineers to a customer under contract rather than the customer hiring their own. It shows up in enterprise AI deals where the platform is strong but adoption is the risk, and increasingly as a standalone offering from consultancies. For a candidate the distinction matters — in a vendor FDE role you are deepening one product; in an FDE-as-a-service role you are closer to high-end consulting, with more variety and less product ownership.

FDE vs Solutions Architect vs Sales Engineer

This is the question that sends most people to this page, and it has a measurable answer. Every posting on this board is scored on seven dimensions read from the description rather than the title. Two of them — how much you code, and how much you sell — separate the family almost completely. Medians from 1,150 live postings, 17 August 2026:

Median score per dimension, 0–100, from the posting text. Full seven-dimension breakdown in the analysis linked below.

Two conclusions worth carrying away. Solutions Architect is closer to Sales Engineer than to FDE — the opposite of what “architect” suggests to most engineers reading the posting. And Applied AI Engineer is nearly the same shape of job as FDE; the difference is subject matter, not structure, with Applied AI roles far more likely to involve models directly.

One caution the medians hide: the spread inside a single title can be wider than the gap between titles. Of postings titled “Solutions Architect”, 40% score 70 or above on sales involvement while 13% score 30 or below — two different careers sharing a phrase. FDE is the most consistent title of the five. Read the full comparison, with all seven dimensions →

What the job requires

The hiring bar is genuinely a software engineering bar, with three additions:

• Breadth over depth. You will touch data engineering, backend, a little frontend, cloud infrastructure and authentication in the same month. Python and SQL are close to universal in these postings; cloud platforms and data tooling — AWS, Azure, GCP, Spark, Databricks — dominate the stack requirements.
• Speed to something working. Forward deployed teams are judged on demonstrated progress in weeks. Engineers who need a complete specification before they start tend to struggle regardless of raw ability.
• Composure with customers. Not charisma — the ability to sit in a room where something is broken, absorb the frustration, and keep asking useful questions.

AI exposure is now common but not universal in this role: 49% of FDE postings scored 70 or above on AI/LLM usage in the August sample, against 96% of Applied AI Engineer postings. If working directly on models is the point for you, that second label delivers it far more reliably.

Who hires Forward Deployed Engineers

Three groups, with different versions of the job. AI labs — OpenAI, Anthropic, Cohere — where the work is getting frontier models into enterprise workflows, and the product is changing under you. Data and enterprise platforms — Palantir, Databricks, Snowflake, Elastic, MongoDB — where deployments are larger, longer, and more regulated. And a long tail of AI startups where “FDE” means you are the deployment organisation, alone.

Because the title space is so fragmented, searching by title alone misses a large share of the market. This board classifies postings by what the description says the job is rather than by what it is called, which is the only way to see the whole set.

What Forward Deployed Engineers earn

From advertised US ranges on this board in August 2026, deduplicated so each employer’s band counts once: the median advertised midpoint for an FDE was about $215,000, with a 10th-to-90th percentile spread of roughly $159,000 to $294,000. That places FDE at the top of this job family alongside Solutions Architect, and roughly $38,000 above Implementation and Delivery roles at the median.

Two caveats matter more than the headline. Only a minority of postings publish a range at all, and the ones that do are concentrated — a single employer can account for most of a role’s disclosures. And the same title pays about $48,000 more at an AI lab than at an enterprise platform, so the role label predicts your pay far less than the kind of company does. See the full salary report, with the disclosure problem handled honestly →

How to get into the role

Most FDEs arrive from one of three directions: a backend or data engineering role, where the gap to close is customer exposure; consulting or technical delivery, where the gap is engineering depth; or sales engineering, where the gap is production ownership. All three are normal, and hiring managers in this family are used to reading non-linear backgrounds.

The loop is where people are most often surprised. It is not a standard backend loop: there is usually a prototyping or take-home round and an ambiguous customer-scenario round, and between them they decide most offers. Strong engineers lose the offer by treating the scenario round as a soft-skills formality. The FDE interview guide covers what each stage is scoring →

Common questions

Browse open roles

Every posting on this board is classified by what the description says the job is, so the role filters catch the work that hides under other titles.

Browse Forward Deployed Engineer jobs

Or start from Applied AI Engineer jobs, Solutions Architect jobs, Solutions Engineer jobs, or Implementation Engineer jobs.