AI Careers

What Is a Forward Deployed Engineer?

The forward deployed engineer role explained: where the title came from at Palantir, what FDEs do, the skills required, and how to build proof for the job.

Mauricio Esparza By ·Published ·6 min read
mithub.club
Short answer

A forward deployed engineer is an engineer who embeds with a customer and owns getting software to produce a real result inside that customer's environment. The title comes from Palantir, where the role is called Forward Deployed Software Engineer and involves embedding engineers directly with customers to own end-to-end delivery of high-stakes projects.

A forward deployed engineer (FDE) is an engineer who embeds with a customer and owns getting the software to produce a real result inside that customer's environment. Not a demo, not a recommendation, not a handover document: a working system in a messy real business, with the engineer accountable for whether it works. The title comes from Palantir, and it has spread across the AI industry for one reason — the hard part of AI is no longer building the capability. It is deploying it.

In short

  • FDEs write code and sit with the customer. Both halves are the job.
  • The title originated at Palantir, where the role is Forward Deployed Software Engineer and was known internally as Delta.
  • The role is growing because deployment, not capability, is the bottleneck in enterprise AI.
  • The skill stack is unusual: domain fluency, diagnosis, integration engineering, measurement and handoff.
  • You can build proof of these skills before anyone gives you the title.

Where the title comes from

Palantir created the role. In The Pragmatic Engineer, Gergely Orosz reports that Palantir established the forward deployed engineer position in the early 2010s, originally calling it "Delta," and that hiring for the role has since expanded to companies including OpenAI, Ramp and Salesforce. The reason it was invented is instructive: powerful software kept failing to turn into operational results at the customer, so the company started sending engineers to the customer instead of sending documentation.

Palantir's own current Forward Deployed Software Engineer posting describes the model as embedding engineers directly with customers to tackle their most pressing challenges, and asks the engineer to own the end-to-end execution and implementation of high-stakes projects: collaborating on architecture, managing large-scale data with AI, developing custom applications, and engaging directly with customer stakeholders. The stated requirements are notably plain — at least one year of relevant post-college experience, strong coding in a language such as Python, Java, C++ or TypeScript, an analytical approach, and the ability to work in teams of technical and non-technical people.

On a typical week, Orosz reports FDEs splitting their time between working at customer premises, meetings with stakeholders, and monitoring, debugging, deploying and configuring software — plus pushing fixes back into the core product when it gets in the way of the customer's implementation.

That last part is the quiet superpower of the role. The FDE is the only person who sees, at the same time, what the product promises and what the customer's reality does to it.

Why the role exploded with AI

Every company now has access to roughly the same models. Almost none of them have their processes, data and permissions in a state where those models can produce revenue.

The gap is not intelligence. It is that a model does not know your pipeline stages, does not know that two of your fields contradict each other, does not know which branch manager will refuse to use a tool, and cannot tell that your compliance constraint makes the obvious design illegal. Someone has to sit inside the business and close that distance — writing the integration, mapping the real process, choosing where a human must stay in the loop, and proving the result with a number.

That job description is the forward deployed engineer. It is also, almost exactly, what MitHub teaches in the Revenue Reverse Engineering faculty: trace the money backwards, diagnose, prove value fast, operate, publish the case study.

FDE vs the roles it gets confused with

RolePrimary outputWho owns the resultTypical failure mode
Forward deployed engineerWorking system in the customer's environmentThe FDEBuilds what was requested instead of what produces the outcome
Solutions / sales engineerA convincing demo and technical answer in the sales cycleSalesDemo works, deployment does not
ConsultantAnalysis, recommendations, a planThe clientGreat plan, nobody implements it
GTM engineerRevenue systems inside their own companyThe GTM teamOptimizes tooling, not pipeline
Platform / product engineerOne capability used by many customersProductShips features no customer operationalizes

The cleanest distinction is scope of loyalty. A product engineer builds one capability for many customers. A forward deployed engineer builds many capabilities for one customer. Everything about the role — the ambiguity, the travel, the direct exposure to whether it actually worked — follows from that inversion.

The MitHub FDE Stack

Five layers. Most people who want this job have one or two and assume the rest are soft skills. They are not; they are the job.

1. Customer fluency. Learn a domain fast enough to challenge it. You should be able to describe how the customer makes money, what their week looks like, and which three numbers their boss is judged on — within your first week. Without this, you build technically correct things nobody uses.

2. Diagnosis. Map the real process, with numbers, and find the constraint before proposing anything. This is the discipline in process mapping before automation, and it is what separates an FDE from a contractor taking orders.

3. Integration engineering. APIs, webhooks, authentication, data shapes, idempotency, retries, error handling. Not glamorous. It is where deployments actually die — usually at the seam between two systems, at 4pm on a Friday, on the 3% of records that are malformed.

4. Measurement. Define the number before you build, instrument the system so every run leaves a signal, and be willing to report that the number did not move. An FDE who cannot prove the result is a contractor with a laptop.

5. Handoff. Document, train, and make yourself unnecessary at this account so you can go be necessary at the next one. The instinct to become indispensable is the single most common career-limiting move in this role.

Underneath all five is one trait that shows up in every FDE job description in some form: tolerance for ambiguity. You will routinely be the only technical person in a room full of operators, with an underspecified problem and a customer who cannot articulate what they want until they see something wrong.

Where the FDE sits on MitHub's ladder

MitHub describes capability in four steps: Doer → Director → Designer → Owner, explained in Director vs Doer.

A forward deployed engineer operates as a Designer — building the systems and workflows — while being held to an Owner's standard, because they are accountable for the business result in someone else's building. That is why the role pays and travels well: it is one of the few jobs where a single person is measured on an outcome, not on output.

How to build proof before you have the title

You do not need permission to practise this. You need one real environment and three artifacts.

  1. Find a real operator with a real problem. A small business, a nonprofit, a friend's agency, your current employer's ugliest internal process. Real beats hypothetical every time.
  2. Produce a diagnosis document. The current process, mapped, with numbers and the constraint identified. This alone puts you ahead of most applicants.
  3. Build the smallest system that moves one number. Integrated with their real tools, logging every run, with a human approval where the mistake would cost something.
  4. Measure for 30 days and report honestly, including what did not work.
  5. Write it up as a case study — situation, path, result, evidence — using the structure in Your case study.

Three of those and you can hold a conversation with any hiring manager for this role, degree or no degree.

What makes FDEs fail

  • Building what was asked instead of what produces the result. The customer's request is a hypothesis, not a specification.
  • Skipping the baseline. Without a before, there is no after, and your work becomes a matter of opinion at renewal time.
  • Going native. You are there to change how the work happens. If you end up doing the work by hand because it is faster, the deployment has failed quietly.
  • No documentation. Knowledge stuck in one head is a system failure, not job security.
  • Treating a green run as a result. The workflow executing successfully is not the same as the business result appearing in the customer's system. Go look at the record.

On pay

Compensation for FDE roles varies widely by company, level and location, and public aggregate figures for a title this new are unreliable. The honest advice: read the ranges published on actual postings at the companies you are targeting, and compare against adjacent roles — see GTM engineer salary for how to research this properly rather than trusting a single number.

Where to go next

The adjacent roles share most of this stack, so read them before you pick a lane — the closest is how to become a GTM engineer, which covers the same diagnosis-and-integration work aimed at a company's own revenue systems. Then stop reading and go build the first artifact: Learn, build, prove, earn.

Frequently asked questions

Where did the term forward deployed engineer come from?

From Palantir, which created the role and called it Delta internally. The Pragmatic Engineer dates its creation to the early 2010s. Palantir's current job postings use the title Forward Deployed Software Engineer.

Is a forward deployed engineer the same as a consultant?

No. A consultant usually recommends and hands over a plan. An FDE writes the code, deploys it in the customer's environment and is measured on whether the system produces a result.

Do you need a computer science degree to be an FDE?

Postings vary. Palantir's own FDSE posting asks for at least one year of relevant post-college experience and strong coding ability in a language such as Python, Java, C++ or TypeScript, rather than a specific degree.

How much do forward deployed engineers earn?

It varies widely by company, seniority and location, and public aggregates are unreliable. Read the compensation range on actual postings from the companies you are targeting.

Sources

  1. Forward Deployed Software Engineer (job posting) — Palantir Technologies (accessed 2026-09-17)
  2. What are Forward Deployed Engineers, and why are they so in demand? — The Pragmatic Engineer (Gergely Orosz, 12 August 2025) (accessed 2026-09-17)
AI CareersForward Deployed EngineerImplementation
Mauricio Esparza
Mauricio EsparzaGTM Systems Lead · Revenue Engineer · Founder of MitHub. Designs and runs revenue systems for multi-location businesses: AI voice campaigns, enrichment, CRM automation and attribution. Founded MitHub to teach the method in the open.

Part of AI Careers on MitHub.

Keep going

GTM Engineering

How to Become a GTM Engineer: Skills, Stack and a 90-Day Plan

A practical path to becoming a GTM engineer: the skills that matter, the tools to learn, a 90-day plan and three proof-of-work projects that companies notice.

Read · 7 min →mithub.club
GTM Engineering

What Does a GTM Engineer Do? Responsibilities, Tools and a Real Workflow

What a GTM engineer does day to day: the build loop, typical projects, tools and skills from job-postings data, pay context and a build ticket to use.

Read · 6 min →mithub.club
AI Careers

AI Implementation Specialist: The Role Behind Real AI Adoption

Buying AI is easy; getting a team to rely on it is not. What an AI implementation specialist does, the adoption ladder, the skills and how to prove them.

Read · 7 min →mithub.club
AI Careers

What Does an AI Automation Engineer Do?

An AI automation engineer builds workflows and agents that run unattended in a business: the real scope, a production checklist, the stack and the path in.

Read · 7 min →mithub.club
AI Careers

The Skills That Make You More Valuable in the AI Economy (and How to Stack Them)

Which skills labor-market data says are rising, why combinations beat single skills, and a MitHub framework to stack AI, business and judgment skills.

Read · 6 min →mithub.club
Career Development

How to Build a Portfolio Without Experience (Revenue and AI Automation Edition)

Build a portfolio with no job history: pick real business problems, build practice systems in revenue and automation work, and document proof others can check.

Read · 6 min →mithub.club