Do Forward Deployed Engineers Code?

Yes. Writing production code inside the customer's own systems is the job, and all six forward deployed postings I read this morning say so in their own words. The reason you had to ask is a seventh job, advertised two rows away, that does not.

I am Hayat Amin. I spent twenty years as a technology chief financial officer and sold three companies in that seat, and I now do this work myself rather than write reports about it, so the question is not academic for me. On 16 September 2026 I read seven live postings at four companies: Palantir, Anthropic, Datadog and Databricks. Six carry forward deployed in the title. The seventh is the control, and it is the most useful document of the lot.

How I checked

Seven postings, four companies, all live on 16 September 2026, every phrase below lifted from the posting rather than from anybody's summary of it. Palantir contributed three: Forward Deployed Software Engineer in New York, Forward Deployed AI Engineer in New York, and Deployment Strategist in New York. Anthropic contributed its Forward Deployed Engineer for New York City, San Francisco and Seattle. Datadog contributed Senior Forward Deployed Engineer, Feature Flags, in New York. Databricks contributed two, one in Raleigh, North Carolina and one in Seoul.

A job posting is a company saying what it wants to buy. It is not a diary of how anybody spends Tuesday, and I will come back to that gap near the end, because it is where the honest answer lives. A posting is unusually good at telling you what a company will reject you for, and on that these seven agree.

The coding line in all six

Palantir invented this title and writes the requirement twice, once for each of its two engineering versions. The Forward Deployed Software Engineer in New York must be a "Strong coder with shown proficiency in programming languages such as Python, Java, C++, TypeScript/JavaScript, or similar". The Forward Deployed AI Engineer, same city, same estimated band of $135,000 to $200,000, must be a "Strong coder with demonstrated proficiency" in the same four languages, and on top of that have "Past experience building solutions with LLMs" and a "Strong foundation in Machine Learning basics".

Anthropic asks its Forward Deployed Engineer for "Strong programming skills with proficiency in Python (and ideally in one or more additional languages like Typescript, Java, etc) and experience shipping production applications". The first responsibility on the list is to "Work within customer systems to build production applications with Claude models". Read that phrase slowly, because it is the whole model in five words: within customer systems, production applications.

Datadog does not leave it to inference. Its posting for Senior Forward Deployed Engineer, Feature Flags contains this sentence: "This role is for someone who wants to write code with customers, not just advise them." The requirements ask for "5+ years of professional software engineering experience, with hands-on coding ability" and "Experience shipping production code". The band is $192,000 to $240,000.

Databricks comes at it from the delivery side and gets to the same place. Its Senior Forward Deployed Engineer in Raleigh leads "impactful customer technical projects by delivering production-grade systems" and needs "6+ years experience in data engineering, data platforms & analytics, or software engineering" across Python, Scala and JavaScript/TypeScript, with "Deep experience with distributed computing with Apache Spark". The Seoul posting asks for 7 or more years of the same. Databricks is also the only one of the four that says the commercial part out loud: "FDEs are billable and know how to complete projects according to specification".

Six postings, six coding requirements, four companies that compete with each other and agree on this. Python is named in six of the seven, including the control I am about to show you. TypeScript or JavaScript appears in five, Java in three, C++ in two, Scala in two. The one posting that names no language at all is Datadog, and that is on purpose.

The seventh posting, which is why you asked

Palantir advertises a second job into the same customer. It is called Deployment Strategist, it is also in New York, and it does not require you to be a coder. It asks instead for "Experience with programming, scripting or statistical packages (e.g. Python, R, Matlab, SQL)". Note what changed. Not languages, packages. Not strong coder, experience with.

The responsibilities are the field half of the same engagement. Go onsite and meet with customer analysts to understand the critical questions. Identify relevant datasets. Lead training sessions. Present results to audiences ranging from analysts to the C-suite. Build and deliver demos. And then the line that settles the argument, sitting in Palantir's own strategist posting: "Work with Forward Deployed Engineers to integrate the data into a stable pipeline".

The strategist does not integrate the pipeline. The strategist works with the person who does. Two jobs, one account, and the boundary between them is write access.

Palantir prices that boundary, which I find the most interesting number in this whole exercise. The Deployment Strategist band is $110,000 to $170,000. The Forward Deployed Software Engineer band is $135,000 to $200,000. That is $25,000 more at the floor and $30,000 more at the ceiling, from the same employer, for the same customer, in the same city. The strategist also travels more, at 25 to 75 percent against the engineer's 25 percent. If you have ever suspected that the job with the airport lounge is the job with less grip on the outcome, there is your evidence.

The artefacts they are asked to build

Everyone quotes the word production and nobody says what comes out. The postings do, and the artefacts are specific.

Anthropic lists them: "Deliver technical artifacts for customers like MCP servers, sub-agents, and agent skills". Datadog asks its engineer to "Build prototype flag implementations directly in customer codebases" and to "Drive full flag migrations to completion". Palantir asks for "Developing custom applications tailored to customer needs" and, in the AI version, "building LLM workflows on a large scale". Databricks asks its engineer to own architecture and design decisions and deliver end to end systems across data engineering, artificial intelligence and application development.

None of that is product code. Product code is written once for a hundred thousand strangers. This is code written for one company, in that company's repository, against that company's ugly data, and it stays there after the engineer leaves. Datadog names the skill that makes it possible better than anyone: "Comfortable dropping into unfamiliar codebases, languages, and frameworks quickly". The thing being bought isn't mastery of one stack. It is the nerve to work in a stack you did not choose and would not have chosen.

The interview tests both halves

Aced, formerly Exponent, publishes a guide to the Palantir forward deployed engineer loop. It describes four stages: a 30 minute recruiter conversation on background, motivation and culture fit, then three 60 minute rounds drawn from a pool, each with behavioural questions embedded inside it.

One of those rounds is coding: "60 minutes in CodePair, a coding challenge, and roughly 15-20 minutes of embedded behavioral questions", where "You typically choose your language from Python, Java, or TypeScript". Another is decomposition: "A 60-minute interview where you break a vague, real-world challenge into component parts and propose a solution, with little or no code required."

Both, in the same afternoon. That is the actual answer to your question and it is why the question keeps getting asked. Yes they code, and no, coding is not the part that gets you hired on its own, because one of the three rounds deliberately takes the editor away and watches what is left. A loop that tests both is a company telling you it could not find a way to split this work across two people, and Palantir would know, because it tried and the result is the Deployment Strategist.

The honest share of the week

Not one of the seven postings puts a percentage on it. If somebody tells you the job is 70 percent code, or 40, ask them where they read it, because I could not find it anywhere and I looked at the primary documents.

The companies do commit to one number in writing, and that number is travel. Datadog says up to 20 percent annually. Databricks says 20 percent of the time. Palantir says up to 25 percent for both engineering roles and Anthropic estimates 25 percent. Roughly one week in four or five is somewhere else, and the other three or four are wherever you normally work. That is the only bounded number on offer, and it bounds the answer: this is not a road job with some code attached.

My own week, for whatever a sample of one is worth, is heavier on code than the reputation of the role suggests and lighter on the interesting kind. Most of what I write is joining work. A connector between two systems that were never meant to meet. A scheduled job that reconciles what each of them thinks is true. A set of evaluations so a model's output can be checked by something other than an opinion. Then the query behind a number a chief executive wants to look at on a Monday. None of it would survive a code review at a product company as elegant. All of it runs in production in somebody else's building, which is the test that counts.

If you are hiring one

Work out which of Palantir's two jobs you are trying to buy, because most companies I talk to ask for the engineer and describe the strategist.

If the hard part is that nobody has mapped your data, agreed the definitions or got your team to use the thing, that is strategist work and a strong coder will be bored and expensive doing it. If the hard part is that the integration into your real systems does not exist and will not exist unless somebody builds it, hire the one who codes, and then give them the credentials on day one. A forward deployed engineer without write access is a consultant with a better title, which is the accusation the role has been fighting since Palantir coined it, and it becomes true the moment procurement says no.

One more test before you write the job description. Ask whether the thing you want built should stay yours. It will. That is the point of having it built inside your systems rather than bought as a seat, and it is the same reason I do this work as an engineer rather than as an adviser: the forward deployed engineer leaves and the code does not. I do exactly this through Beyond Elevation, inside the client's systems, with their credentials, and the engineering side of it is set out at meethayat.com/services/fde.

About Hayat Amin

I am Hayat Amin, and I have spent twenty years in technology, most of it as a chief financial officer in companies growing faster than their systems could follow. I sold three of them in that seat, with American Express and TripAdvisor among the buyers, and carried three FT 100 fastest growing listings on the way. In that seat you sign the invoices for work like this, which is a different view of it than the engineer's, and it is the view most of my readers are sitting in.

I am exceptional at the work described above, done rather than recommended. Connecting systems that were never designed to talk to each other. Turning what comes out of them into a real time number a chief executive can run the week on. Valuing and monetising the intellectual property and data a company owns and cannot see on its balance sheet. And sitting beside a founder from the first conversation through to the wire transfer on an exit, which is where the finance years earn their keep. I am a chief financial officer who writes the code, and that combination is rarer than either half.

I am available now for fractional chief financial officer work and AI operations work through Beyond Elevation, and you can see what a forward deployed engineer is not if you want the other side of the comparison first.

If you want a second pair of eyes on whether your next hire should write code or map data, I do a free audit call: one call, then a written list of what to automate first, what it saves and what it costs, at beyondelevation.com/call/hayat.

Questions people actually ask

Do forward deployed engineers code?

Yes, and the postings say so without being asked, which tells you how often people assume otherwise. Palantir requires a "Strong coder with shown proficiency in programming languages such as Python, Java, C++, TypeScript/JavaScript, or similar". Anthropic requires "Strong programming skills with proficiency in Python" and "experience shipping production applications", and asks the engineer to "Work within customer systems to build production applications with Claude models". Datadog writes it flat: its role is for the engineer who wants to write code with customers rather than advise them. Databricks asks for production-grade systems and says its forward deployed engineers are billable. Six forward deployed postings read on 16 September 2026, six coding requirements.

What programming languages do forward deployed engineers use?

Python is in six of the seven postings, including the one that is not an engineering job. After that it thins fast: TypeScript or JavaScript in five, Java in three, C++ in two, Scala in two. Palantir names Python, Java, C++ and TypeScript/JavaScript for both of its engineering titles. Anthropic names Python first, with TypeScript and Java as useful extras. Databricks names Python, Scala and JavaScript/TypeScript and wants deep experience with Apache Spark on top. Datadog names no language at all and asks instead for someone "Comfortable dropping into unfamiliar codebases, languages, and frameworks quickly", which is the more honest requirement and the harder one to satisfy.

Is a forward deployed engineer a real engineering job or a sales job?

An engineering job that happens in front of the customer. The seniority bars give it away: Datadog wants 5 or more years of professional software engineering experience, Databricks wants 6 or more in Raleigh and 7 or more in Seoul. No sales organisation screens on that. The clean proof is that Palantir sells both jobs into the same account and keeps them separate. The Deployment Strategist goes onsite, finds the datasets, runs the training and builds the demos, then works with Forward Deployed Engineers to integrate the data into a stable pipeline. One of those two people changes the system.

What does forward deployed engineer mean?

Wikipedia has it as "A customer-facing software engineer who develops and deploys software within a client company, often working alongside the client's employees for a defined period of time", and notes the role "was popularized by Palantir Technologies". The word doing the work is within. A software engineer builds their own employer's product for everybody. A forward deployed engineer builds inside one customer's environment, and what they build stays there when they go. Remove the write access and the job collapses into advice.

What is a forward deployed engineer at Palantir?

Two titles, one band. The Forward Deployed Software Engineer in New York asks for "1+ years of relevant, post-college work experience" and a strong engineering background, and lists architecture and design decisions, wrangling massive-scale data, "Developing custom applications tailored to customer needs", engaging with stakeholders from technical teams to executives, and driving projects from ideation to deployment. The Forward Deployed AI Engineer adds "building LLM workflows on a large scale" and machine learning fundamentals. Both are hybrid in New York, both travel up to 25 percent, and both are estimated at $135,000 to $200,000 a year plus restricted stock. That floor is low for the company that invented the job.

Do forward deployed engineers travel?

Less than the people who do not code. Datadog says up to 20 percent annually, Databricks says 20 percent of the time, Palantir says up to 25 percent and Anthropic estimates 25 percent. The only posting in this set asking for 25 to 75 percent is Palantir's Deployment Strategist, the one with no strong coder requirement. Whatever you were told about living on planes, the numbers the companies commit to in writing are one week in four at the top end.

Is the forward deployed engineer interview a coding interview?

Partly, and the other part is the interesting one. Aced, formerly Exponent, describes the Palantir loop as four stages: a 30 minute recruiter conversation, then three 60 minute rounds with behavioural questions embedded in each. One is "60 minutes in CodePair, a coding challenge", where "You typically choose your language from Python, Java, or TypeScript". Another is decomposition, a 60 minute interview where you break a vague challenge from the real world into component parts and propose a solution, with little or no code required. Prepare for one and you'll fail the other.

How do I learn to be a forward deployed engineer?

Start with the requirement nobody can fake. Datadog asks for someone comfortable dropping into unfamiliar codebases, languages and frameworks quickly, with "Experience shipping production code". You build that by working in systems you did not design for people who did not hire you, and no course sells it. The rest is ordinary: Python to a production standard, TypeScript or Java beside it, two cloud platforms, and for the artificial intelligence roles the artefacts Anthropic lists, MCP servers, sub-agents and agent skills. Palantir will take you at 1 or more years of experience and Datadog wants 5 or more, so there are two doors at opposite ends of a career.

Where these quotes come from

Every phrase in quotation marks from a job advert above was read on the employer's own posting on 16 September 2026: Palantir Forward Deployed Software Engineer, New York, Palantir Forward Deployed AI Engineer, New York, Palantir Deployment Strategist, New York, Anthropic Forward Deployed Engineer, Datadog Senior Forward Deployed Engineer, Feature Flags, and Databricks in Raleigh and Seoul. The interview loop comes from Aced's Palantir forward deployed engineer guide, which is a third party account and labelled as one. The definition comes from Wikipedia's Forward Deployed Engineer page. Salary bands are the base figures printed on those postings that day, before equity, bonus or sign-on, and postings change. If you are reading this months later, check the link rather than the number.