What is a forward deployed engineer? The FDE role, explained

In this lesson you'll learn:

• what an FDE actually does, in the words of the companies that hire them;
• the job's three stages, Frame, Deliver and Evolve, and how the role differs from a solutions architect;
• who works beside an FDE, and why some weeks nobody does.

Sound familiar?#

• The model works in the demo and nowhere near the customer's systems, data or approvals.
• Every enterprise deal ends in a custom build nobody at the vendor can maintain.
• The product team learns what customers need from sales notes, months late.

The FDE exists because all three are the same problem: the distance between a capable product and a working deployment.

What does a forward deployed engineer do?#

Writes production software inside a customer's environment, and owns the whole path to value. Palantir, which created the role, framed it against its product engineers: a product engineer's focus is "one capability, many customers", while a forward deployed engineer's is "one customer, many capabilities". Six companies that run FDE teams describe the modern version in their own words:

Two threads run through the rows: the FDE ships to production, not to a demo, and the FDE's second customer is their own product team. Ramp's and AWS's rows add the span of the job: from a prospect in a sales funnel to a customer whose own engineers run the system.

How the role works, stage by stage#

In three stages, named so that their initials spell the role: Frame, Deliver, Evolve. Each asks the manual's four questions, P0 to P3, about a different thing: the engagement, the system, the relationship. Each ends with a signature on the customer's side. The forward-deployed engineer guide gives every stage four steps, each with its template, its prompts and a worked example.

Stage 1 · Frame the engagement#

Days, before a contract or a charter. Measure the pain in the customer's data, not the pitch: cases, minutes and money, the systems the agent must touch, and who owns the risk. Then write a statement of work a sceptic would sign, prove it on their cases in days, and read the proof out with the failures first. An FDE says no more than most engineers: to the flagship first feature, to building past an unsigned autonomy decision, to meetings that produce no artefact. The stage ends with the go decision, signed by the customer's sponsor. Frame the engagement

Stage 2 · Deliver the system#

Weeks. Start every lead-time item on day one (model access, data access, the security review), because they set the calendar. The code lives in the customer's repository, runs in their environment and uses their identity model. Their engineers pair on it, their experts label the golden set, their risk owner signs every limit, and a shadow run compares the agent with their staff: an FDE's claims are the customer's numbers, not the vendor's. The stage ends with the handover, signed by the person who will run it. Deliver the system, and the stage in depth: AI-DLC and AIDD in the field

Stage 3 · Evolve the relationship#

Months, for as long as the system runs. Turn what the running system taught into the next frame, and write the patterns up for the people who build the product: the connector built for the third time, the evaluation that exposed a model gap. Prove the reusable version at the next deployment, and hold a value review each quarter, the saving beside the spend. The stage ends with the next frame, or a clean close, signed by the customer's sponsor. Evolve the relationship

How is an FDE different from a solutions architect?#

The distinction that matters most is the last column combined with the feedback duty. A consultant can succeed with bespoke work; an FDE who only ever builds bespoke work has turned a product company into a services firm. OpenAI's own framing, as reported, distinguishes its FDEs from the professional-services work of a solutions architect.

Who works beside an FDE?#

At a larger vendor, a partner; some weeks, nobody. At a customer you may be the only person from your side, and then you do every job: you size the pain as a product manager would, draw the system as an architect, build it as an engineer, prove it as a QA lead, keep it running as a platform team and lead the room as a consultant.

At scale the work is a pair. The FDE owns the how; a partner owns the what and the why, and each company that runs FDE teams names the partner differently (postings read on 2 October 2026):

The partner's hardest call is sorting every customer request into configuration, a service or a product capability, and defending that call to the customer and to the product team. Learn both halves: when the pair is one short, you are both.

Where you'll use it#

• When hiring or joining an FDE team, to know which of the three stages the role really owns, and who works beside it.
• When scoping an enterprise AI deal, to decide what the FDE owns and what the customer must.
• When an engagement drifts into bespoke work, to put the product feedback duty back on the plan.

Why it matters#

Models are capable; deployments are hard. Most of the distance between the two is integration, access, evidence and trust: work that is specific to one customer and cannot be done from a distance. The FDE is how a product company closes that distance without becoming a consultancy, provided the patterns flow back.

Key takeaways#

1. An FDE owns production outcomes at one customer, in three stages: frame, deliver, evolve.
2. The second customer is the product team: repeated work becomes a pattern, then a product.
3. FDEs say no by design: to the flagship first, to unsigned autonomy, to bespoke forks.

FAQ#

What does FDE stand for?#

Forward deployed engineer: a software engineer who works embedded with a customer to deploy and adapt a product in the customer's environment. Palantir's original title was forward deployed software engineer.

Is a forward deployed engineer a software engineer?#

Yes. FDEs write production code, usually in the customer's systems and stack, but the job also includes discovery, scoping, stakeholder work and evaluation, so the balance of the week differs from a product engineer's.

What skills does a forward deployed engineer need?#

Strong programming, fast learning of unfamiliar systems, discovery with non-technical stakeholders, evaluation design for AI systems, security and identity basics, and the judgement to say no to work that cannot be proven or maintained.

What is the difference between an FDE and an FDPM?#

The FDE owns how a capability is built and deployed at a customer; the forward deployed product manager owns what is built and why, and decides which customer needs become product capabilities.

Apply it in your role#

Across the enterprise. Run FDEs as a programme with a feedback budget: a fixed share of every engagement goes to codifying patterns, and the product roadmap has a standing slot for them.

The ten-minute workflow. Keep the pattern log honest:

Here are my notes from this week's customer work: <paste>. Extract every problem I solved, and mark the
ones I have solved before at any customer (check against this log: <paste the log>). For each repeat,
write a pattern entry: the problem, the fix, the customers, and whether it should become configuration,
a reusable service or a product capability.

Sources and credits#