What Is a Forward Deployed Engineer?
A forward deployed engineer is a senior engineer who stops building inside their own company and starts building inside yours. Write access to your systems. Production code in your environment. Still accountable after the invoice clears. That is the whole job, and everything else written about it is decoration.
I could give you another paragraph of definition. Instead I went and counted. On 12 September 2026 I pulled every open role from the public careers feed of ten companies that run this function, read the postings with forward deployed in the title, and wrote down what each one actually asks a human being to do. Three thousand seven hundred and twenty three open roles. Two hundred and forty of them forward deployed. The definition below is not mine. It is theirs, and they broadly agree, which is more than you can say for most job titles invented in the last two years.
How I built this list
The method matters more than the ranking, so here it is in full, and you can repeat it yourself this afternoon.
Ten companies, chosen because each one has a public, machine readable careers feed and at least one live posting with forward deployed in the title. For each, I took the complete list of open roles on 12 September 2026, counted the forward deployed ones, and read the postings in the United States wherever a United States posting existed. The order below is by the number of open forward deployed roles, because that is the one number that is measured rather than judged. Every salary figure is the base band printed on the company's own posting that day, before equity, bonus or sign-on. Every phrase in quotation marks is lifted from the posting itself.
What this list is not: a ranking of who is best to work for, which I cannot know, or a headcount, which none of them publish. Open postings measure hiring intent in September 2026. That is a real signal and a narrow one, and I would rather tell you the edge of it than pretend the number is bigger than it is.
Two companies I expected to include are missing and it is worth saying why. Sierra had 210 open roles and not one with forward deployed in the title, despite the founder's well known public argument for the model, so it is not in the count. Google Cloud and Deloitte both run this function, and the figures I have for them, 59 forward deployed hires at Google Cloud in 2026 and a Deloitte service page with 19 open postings, come from reporting rather than from a feed I read this week. They are context, not entries.
1. Databricks: 95 forward deployed roles out of 886
The largest forward deployed hiring programme in the sample, by a distance, and the clearest statement of what the job is when it lives inside a delivery organisation. Ten point seven percent of everything Databricks is hiring for right now is forward deployed.
The Senior Forward Deployed Engineer posting in New York City asks the engineer to own the architecture, lead design decisions, and implement end-to-end systems spanning data engineering, AI, and application development. The band is $182,000 to $250,208, and unusually it is identical across all four of the company's pay zones, so geography does not move the number. Travel is twenty percent.
What it is good at: volume and structure. The roles are cut by vertical, fourteen in communications, media, entertainment and games, nine in financial services, seven each in healthcare and life sciences and in retail, four in manufacturing. If you want to learn this job in a system that has already worked out how to run it, this is the deepest bench there is.
Who it is wrong for: anyone who does not want to be billable. Databricks says it in the posting, in its own words: FDEs are billable and know how to complete projects according to specification. That is professional services with a better title and a better platform, and it is honest about it. If you wanted to be shipping product to millions of users, you are in the wrong chair.
2. Palantir: 77 forward deployed roles out of 311
The highest density anywhere. Roughly one in four open jobs at Palantir carries the words forward deployed, which is what happens when a company does not add a function but is built out of one.
Palantir invented the title and says so at the top of the posting: the Forward Deployed Software Engineer role is not just a job title, it is the blueprint, and the company pioneered the position by embedding engineers directly with customers. Its Forward Deployed AI Engineer posting in New York gives the best one line description of the job I have read anywhere: the responsibilities look similar to those of a hands-on AI startup chief technology officer.
What it is good at: range, and the hardest problems on offer. Seventy seven postings covering software, AI, infrastructure, reliability, security and enablement engineering, across commercial work and the United States, United Kingdom, Australian, Japanese, Korean and Norwegian governments and NATO. Alongside them sit 35 Deployment Strategists, whose posting asks them to work with Forward Deployed Engineers to integrate the data into a stable and extensible pipeline. That pairing, a builder and a strategist on the same account, is the original shape of the model and most people copying it copy only half.
Who it is wrong for: people who want to be paid the most, which surprises everybody. The New York Forward Deployed Software Engineer band is an estimated $135,000 to $200,000 plus restricted stock, the lowest floor in this entire list. The company that created the job pays less base than eight of the nine companies copying it. It is also wrong for anybody unwilling to work on defence and government deployments, which are a large share of those 77 roles and are not optional to have an opinion about.
3. OpenAI: 21 forward deployed roles out of 790
OpenAI is the company that split the job in two, and the split is the most useful thing on this page for anyone trying to understand the title.
There are two separate postings in New York. The Forward Deployed Engineer, at $185,000 to $300,000 plus equity, leads complex end-to-end deployments of frontier models in production alongside the most strategic customers, and owns discovery, technical scoping, system design, build, and production rollout. The Forward Deployed Software Engineer, at $153,000 to $325,000 plus equity, works with customers and with OpenAI forward deployed engineers to design abstractions to solve customer problems, and then uses them to scale speed and quality of delivery across all forward deployed engagements. One faces the customer. One turns what the first one learned into reusable machinery. Both are hands on. Both require travel up to fifty percent, on top of three days a week in the office.
What it is good at: geography and specialisation. The 21 roles cover New York, San Francisco, Seattle, London, Madrid, Tokyo, Seoul, Singapore and Sydney, with dedicated healthcare, legal, government and security postings. If you want to see the same deployment problem in nine countries, nobody else is offering that.
Who it is wrong for: anyone with a life that does not survive fifty percent travel plus a hybrid office mandate. That is the heaviest published travel requirement in this list, more than double Databricks. Read that line before the salary line.
4. Scale AI: 19 forward deployed roles out of 223
Scale AI is on this list twice over, because it is the clearest evidence that forward deployment has stopped being one job and become a family of them.
Alongside forward deployed engineers in generative AI and a public sector forward deployed software engineer hiring into Austin, Colorado Springs, Honolulu, St. Louis and Washington, it is hiring three Forward Deployed Product Managers and a Director of Product Management, Forward Deployed and Strategy. The Enterprise posting in New York pays $240,000 to $300,000 and defines itself by exclusion in its first paragraph: this is not a roadmap PM, a CSM, or a solutions engineer. The forward deployed product manager owns product outcomes inside a portfolio of enterprise accounts, drives deployments from contract to production, and distinguishes where the product is the constraint from where execution, integration, or change management is.
What it is good at: the government and frontier agent work, and the honesty of the product manager postings. It also has the widest international spread of the forward deployed engineering roles, including London, Riyadh and Doha.
Who it is wrong for: engineers who assume the product manager version is a softer version of the same job. It is not. It pays more than most of the engineering bands here and asks you to be the accountable person in front of a chief executive while somebody else writes the code. If you want your hands on the keyboard, take the engineering posting, not the product one.
5. Cresta: 10 forward deployed roles out of 94
The smallest company in the top five and the one that shows what the job looks like once the problem is narrow and the product is finished.
Cresta sells conversational AI into contact centres, and its Senior Forward Deployed Engineer posting for the United States, at $185,000 to $235,000, is about deploying agents rather than inventing them: develop, configure, deploy and optimise AI agents on Cresta's platform, build integrations with external systems such as APIs, databases and customer relationship management systems, then optimise performance and troubleshoot in complex enterprise environments. The company says it has raised more than $270 million and names United Airlines, Cox Communications and Marriott as customers.
What it is good at: remote work, which almost nobody else on this list offers for this role. The ten postings cover the United States, Canada, the United Kingdom, Germany, Australia and Taiwan, mostly remote. It also runs forward deployed product managers, including an associate level one, which is the only genuine entry point into this family of jobs I found anywhere in the 240.
Who it is wrong for: anyone who wants variety. This is the same deployment shape, over and over, into contact centres. That repetition is exactly why it can be done remotely and exactly why it will bore a certain kind of engineer by month nine.
6. Anthropic: 6 forward deployed roles out of 599
One percent of Anthropic's open roles are forward deployed, and they pay the highest published base band in this entire list. Both of those facts are the same fact.
The Forward Deployed Engineer posting covering New York City, San Francisco and Seattle pays $280,000 to $320,000 and describes the work with unusual specificity: work within customer systems to build production applications with Claude models, and deliver technical artifacts for customers like MCP servers, sub-agents, and agent skills that will be used in production workflows. Travel is estimated at twenty five percent. The entry requirement is four or more years in a technical, customer facing role, and the posting explicitly encourages former technical founders to apply. It also says the hire will be one of the founding forward deployed engineers, which is to say the function is two years old at the company that is best placed to know whether it works.
What it is good at: price, seniority and the specific artefacts. If you want to know what the AI deployment job will look like in 2028, read what Anthropic is asking for in 2026, because the list of deliverables is concrete in a way most of these postings are not.
Who it is wrong for: almost everybody, statistically. Six openings out of 599. It is a narrow door with a queue behind it, and I would not build a career plan around walking through it.
7. Cursor: 6 forward deployed roles out of 127
Cursor is the entry where the customer is not a business unit. It is another engineering team, and the work is measured in their delivery speed.
The San Francisco posting is the most quotable in the sample: this is not a demo role, you are responsible for systems that work in the real world. The work is described as building Cursor-powered workflows, given as large scale refactors, migrations, pull request review loops, incident to fix pipelines and specification to implementation systems, then owning production quality including tracing, evaluations, metrics, debugging model behaviour and latency and cost tradeoffs. It asks for five or more years of development experience and two or more in a customer facing role.
What it is good at: a tight feedback loop. When your customer is a staff engineer, you find out within a week whether what you built is any good. There is also a Forward Deployed Strategist posting, the Palantir pairing reappearing at a company founded two years ago, which tells you something about how durable that structure is.
Who it is wrong for: anyone who needs a published salary band before they apply. Cursor prints none on these roles, in a list where eight of the other nine companies do. Also wrong for people who are not, themselves, strong engineers. Your customer will know inside one call.
8. Baseten: 3 forward deployed roles out of 94
Baseten runs inference infrastructure, says it recently raised a $1.5 billion Series F led by Altimeter Capital, Conviction Partners and Spark Capital, and names Cursor, Notion, OpenEvidence, Abridge, Clay, Gamma and Writer among its customers. Its Forward Deployed Engineer posting in San Francisco pays $165,000 to $330,000 plus equity.
It is here for one sentence, which is the most honest line about this job that I found in 240 postings: to be clear, this is an engineering role with hands-on coding and software development that also includes aspects of product management, technical customer success, and pre-sales solution engineering mixed in. Every other company on this list is describing that same blend. Baseten is the only one that admits the boundary is deliberately blurred.
What it is good at: hard technical constraints. The work is described in terms of quality, latency and cost outcomes, which are numbers, not adjectives. If you like being measured, this is a measurable version of the job.
Who it is wrong for: anyone who needs a clean role boundary to do good work. Some people are made unhappy by a job that is four jobs. The posting is warning you in advance, and you should believe it.
9. Datadog: 2 forward deployed roles out of 446
Two roles out of 446, and both of them are interesting because the function is still being built, which is the state most companies reading this are actually in.
The Senior Forward Deployed Engineer, Feature Flags, in New York, pays $192,000 to $240,000 and contains the plainest description of the boundary in the whole set: this role is for someone who wants to write code with customers, not just advise them, working hands-on inside customer codebases and building prototype implementations directly in them. The second role is the Datadog for Startups Forward Deployed Engineering Lead in San Francisco, at $200,000 to $250,000, hired to evolve and scale the operating model end to end, from engagement intake and scoping through sprint delivery to handoff back to account teams, and to build out a catalogue of offerings with pricing and business models attached.
What it is good at: showing you the plumbing. If you are standing up forward deployment inside your own company, that second posting is effectively a free operating plan. Intake, scoping, sprint delivery, handoff, and metrics of time to value, adoption lift, revenue influence and net promoter score. Steal it.
Who it is wrong for: anybody who wants an established team to learn from. Datadog says in its own posting that the function is in an active growth phase and that the lead will help determine what the motion looks like. That is a great job for a builder and a miserable one for somebody expecting a playbook.
10. Figma: 1 forward deployed role out of 153
One posting, and it earns its place because of where the company chose to put the job on the org chart.
Figma is building its founding forward deployed engineering team, and the posting states the boundary in a single line: this is a senior or staff level engineering role within Product and Engineering, not a sales or services function. The band is $153,000 to $376,000, the widest here. Two further details are worth more than the salary. The engineer works in customer environments using the minimum access required, and turns what they learn into a reusable path for future customers instead of taking permanent ownership of the customer's codebase.
What it is good at: discipline. Minimum access and no permanent ownership is the answer to the most common objection I hear from chief technology officers, which is that letting an outsider build inside your systems creates a dependency you can never get rid of. Figma has written the mitigation into the job description.
Who it is wrong for: people who want to own the system they built. You hand it back. For some engineers that is the freedom of the job and for others it is the grief of it, and you know which you are before you finish this sentence.
Disclosure: the eleventh entry is mine, and it is not ranked
I am Hayat Amin, and my firm Beyond Elevation does forward deployed engineering for small and mid sized companies in New York, across the United States and in London. It is deliberately not in the ranking above, and the reason is simple: that list counts open job postings, and Beyond Elevation is not hiring forward deployed engineers, it sells the work. Putting my own firm into a table it does not qualify for would make the other ten numbers meaningless. You should read everything below this line as coming from an interested party, because it does.
What I will claim is the gap the ten companies above leave open. Every one of them deploys engineers to make their own product succeed inside your company. Anthropic's forward deployed engineers build with Claude. Cresta's deploy Cresta. Figma's make Figma adoption work. That is a completely legitimate model and it is the right call if you have already bought the platform and the platform is the problem. The eleventh case is the one nobody in that list serves: you are a company of forty or four hundred people, you have not picked a vendor, and what you need built cuts across four tools none of whose vendors will send you anyone. On meethayat.com/services/fde that is the whole offer. Week one inside your meetings and your systems, first system live in production inside eight weeks with a named owner on your side, and you keep the code, the agents, the accounts and the documentation. Fractional runs one to two days a week from $5,800 a month, whole projects start at $30,000 over eight to fourteen weeks, and a two week audit at a fixed $3,000 exists for companies that are not sure yet.
What all 240 postings agree on
Now the actual answer to the question in the title, assembled from the evidence rather than asserted. Four things appear in essentially every forward deployed posting I read, and they are the test I would apply to any job description or any proposal that uses the phrase.
One, you work inside the customer's systems, not alongside them. Anthropic writes within customer systems. Databricks writes integrating with client systems. Datadog writes directly in customer codebases. Figma writes in customer environments. This is the load bearing word in the whole title, and it is the one that separates the job from consulting. Write access or it is advice.
Two, you build, and the artefact is software. Not a deck, not a roadmap, not an assessment. Cursor says it is not a demo role. Datadog says write code with customers, not just advise them. Palantir compares its AI engineers to a hands-on chief technology officer. Baseten calls it an engineering role with hands-on coding, in those words.
Three, you own it through to production, not to handover. OpenAI: own technical delivery across multiple deployments from first prototype to stable production. Scale AI: drive deployments from contract to production, and you know what production-ready actually means. Databricks: lead impactful customer technical projects by delivering production-grade systems. The word production appears in every single posting in this sample. It is the word doing the commercial work.
Four, you go there. Twenty percent at Databricks, twenty five percent at Anthropic, up to fifty percent at OpenAI, twenty five to fifty percent for Anthropic in Europe. The exception is Cresta, which runs it remotely, and notably Cresta has the narrowest and most repeatable deployment of the ten.
And here is the one thing they do not agree on, which is the single most useful sentence on this page if you are buying rather than applying. Half of these companies place the job inside engineering and half place it inside billable delivery. Figma is explicit: not a sales or services function. Databricks is equally explicit in the other direction: FDEs are billable. Same title, two different jobs, two different sets of incentives. Before you hire one or buy one, find out which of the two you are actually getting, because it determines whether the person optimises for your system working or for the hours on the engagement.
The forward deployed product manager is a real job now
This deserves its own heading because it is the fastest moving thing in the data and almost nobody has written it down.
Across those ten careers feeds there were six open forward deployed product manager postings, three at Scale AI and three at Cresta, including an associate level one, plus a director of product management for forward deployed and strategy. A year ago that title barely existed. It is not a forward deployed engineer who cannot code, and Scale AI troubles to say so in the first paragraph: not a roadmap PM, not a customer success manager, not a solutions engineer. The job is to own product outcomes inside a small portfolio of enterprise accounts, to be the accountable escalation point before a problem reaches the customer, and to work out whether a stalled deployment is blocked by the product or by execution, integration and change management, which are three completely different fixes.
For buyers, the useful signal is the price. Scale AI pays that role $240,000 to $300,000 in New York, above most of the forward deployed engineer bands in this sample. Companies do not pay a premium for a translator. They pay it for the person who can stand in front of a chief executive and say the deployment is stuck because of us, or because of you, and be believed. If your own stalled AI project has an engineer on it and is still stuck, that judgement is what you are missing, not more code.
Where the money actually sits
Eleven published base bands, read on the same day, in United States dollars, before equity. Anthropic $280,000 to $320,000. Scale AI's forward deployed product manager $240,000 to $300,000. Datadog's startups lead $200,000 to $250,000 and its feature flags engineer $192,000 to $240,000. OpenAI $185,000 to $300,000 for the engineer and $153,000 to $325,000 for the software engineer. Cresta $185,000 to $235,000. Databricks $182,000 to $250,208. Baseten $165,000 to $330,000. Figma $153,000 to $376,000. Palantir an estimated $135,000 to $200,000.
Two things fall out of that column. The first is that the floors vary by more than a factor of two for a job with the same four requirements, which tells you the market has not settled and that title alone is worth nothing in a negotiation on either side of the table. The second is that the company that invented the role pays the lowest floor in the set, plus restricted stock, and still has the most people doing it. Palantir is selling the experience rather than the salary, and for a decade that trade has worked.
The buyer's version of that maths is straightforward. A forward deployed engineer in New York is a $200,000 to $320,000 base salary, so call it $280,000 to $450,000 fully loaded, and that is before you find one, which is the harder part. That number is the honest benchmark against any quote you receive for the same work, including mine.
When you do not need one
I sell this work, so I will be careful to be useful here rather than persuasive.
You do not need a forward deployed engineer if the work is a product somebody already sells. Buy the product. You do not need one if procurement or security cannot grant an outsider write access to a real system within a month, because the entire value of the role evaporates without it and you will have paid a premium for a very expensive advisor. You do not need one if you have engineers already assigned to the problem and what keeps slipping is sequencing between three teams, which is a coordination problem and is answered in forward deployed engineer vs project manager. And you do not need one if what you actually want is a written opinion. That is a consultant, that is a legitimate purchase, and I have argued the difference at length in is a forward deployed engineer just a consultant.
You do need one when the thing does not exist, nobody on your payroll is going to make it exist this quarter, the hard part is your own systems rather than the model, and you can name the internal person who will own it afterwards. That last condition is the one companies skip, and it is the one that decides whether anything survives the engagement.
What I would do with this if I were you
If you are reading this as an engineer weighing the move, the useful filter is not the salary column. It is the sentence in the posting that tells you whether the job sits in engineering or in billable delivery, and the travel percentage. Those two facts will describe your life more accurately than the title or the band.
If you are reading this as a founder or a chief executive, take the four tests above into whatever conversation you are about to have. Ask the person or the firm across the table whether they will work inside your systems, whether they write the production code, whether they are there after go live, and how much of them you get in the room. Four yeses is forward deployment. Anything less is a different purchase wearing the fashionable word, and the fashion is moving fast enough that plenty of people will use it without meaning it.
I came to this from the finance chair rather than the engineering one, twenty years in high growth companies and three exits, which is why I keep returning to the same question: what is different about the company after the money is spent. On this particular purchase the answer should be a system in production with your name on it, not a report. If you want half an hour on whether your stalled project is a build problem or a coordination problem, that is at meethayat.com/services/fde, and if the constraint is money rather than delivery, the fractional finance seat is at meethayat.com/cfo. The neighbouring questions are forward deployed engineer vs solutions engineer, vs solutions architect and vs sales engineer. The firm side of it is at Beyond Elevation. Otherwise start at meethayat.com.
Questions I get asked about this
What does forward deployed engineer mean?
The phrase is borrowed from the military, where forward deployed means stationed away from home base and close to the problem. In software it means the same thing with commercial force. The engineer is not at headquarters building product for everybody, they are inside one customer's environment building what that customer needs. Palantir, which coined it, calls its version the original forward deployed software engineer and says it pioneered the position by embedding engineers directly with customers. The test that separates it from every neighbouring job is write access. If the person cannot change your system, they are advising you.
Do forward deployed engineers code?
Yes, and every posting read for this piece goes out of its way to say so, which tells you how often people assume otherwise. Cursor: this is not a demo role, you are responsible for systems that work in the real world. Datadog: for someone who wants to write code with customers, not just advise them. Figma: a senior or staff level engineering role within Product and Engineering, not a sales or services function. Palantir compares its forward deployed AI engineer to a hands-on AI startup chief technology officer. Baseten is the bluntest of the ten: an engineering role with hands-on coding and software development, with product management, technical customer success and pre-sales solution engineering mixed in.
Do forward deployed engineers travel?
Yes, and less than the reputation, with one outlier. Databricks says twenty percent for its senior role in New York City. Anthropic estimates twenty five percent in the United States and twenty five to fifty percent for its Munich posting. OpenAI is the heavy one at up to fifty percent, on top of three days a week in the office. Cresta runs most of its forward deployed engineering remotely across six countries, which is possible precisely because its deployments are the most repeatable in this sample. If someone tells you the job is mostly on planes, they are describing 2019 Palantir, not 2026.
What is a forward deployed product manager?
A real and separate job, not a mistyped search. Six open postings across these ten companies on 12 September 2026, three at Scale AI and three at Cresta, plus a director of product management for forward deployed and strategy. Scale AI defines it by exclusion: not a roadmap PM, not a customer success manager, not a solutions engineer. The job owns product outcomes inside a portfolio of enterprise accounts, drives deployments from contract to production, and works out whether a stall is caused by the product or by execution, integration and change management. The New York band is $240,000 to $300,000, above most of the engineering bands here, which tells you how scarce that judgement is.
How many forward deployed engineers does Palantir have?
Palantir does not publish that headcount and anybody quoting you a number is guessing. What is public is the hiring. On 12 September 2026 Palantir had 311 open postings and 77 carried forward deployed in the title, about one in four, the highest density in this sample. They cover software, AI, infrastructure, reliability, security and enablement engineering, across commercial work and the United States, United Kingdom, Australian, Japanese, Korean and Norwegian governments and NATO. Beside them sit 35 deployment strategists, the non-coding half of the same team.
Forward deployed engineer vs software engineer: what is the difference?
Whose system you have write access to, and whether the thing you build is for everybody or for one company. A software engineer builds product in their own repository for a population of users they will never meet. A forward deployed engineer builds inside one named customer's environment against that customer's constraints, and success is measured by whether that one deployment reaches production and gets used. Same craft, different accountability. I have set the two side by side in detail at Beyond Elevation.
What is the forward deployed engineering model?
The model, as opposed to the job, is the decision to solve deployment with your own engineers rather than with a partner network or documentation. Palantir's original shape pairs a builder with a non-coding counterpart on the same account, the deployment strategist, and that pairing keeps reappearing: Databricks has engagement managers and field technical program managers alongside its 95 engineers, Scale AI and Cresta have forward deployed product managers, and Cursor, founded two years ago, already advertises a forward deployed strategist. The pattern is consistent enough to be worth copying. One person holds the relationship and the scope, another holds the keyboard, and they are not the same person because doing both badly is the default failure.
Where do these numbers come from?
Every figure on this page was read on 12 September 2026 from the company's own live careers feed, not from a summary or an aggregator. Databricks, Anthropic, Datadog, Scale AI, Cresta and Figma from their public Greenhouse job boards; Palantir from its public Lever board; OpenAI, Cursor and Baseten from their public Ashby boards. Role counts are of open postings on that date, not of employees. Salary figures are the base bands printed on the postings, in United States dollars, excluding equity, bonus and sign-on, and Palantir's is described in its own posting as an estimate. The hiring growth context, the OpenAI deployment subsidiary, Ode with Anthropic and the Google Cloud and Deloitte figures come from reporting by Pragmatic Engineer, The New Stack, Perspective AI and Paraform rather than from a feed I read this week, and are flagged as such above. Job boards move daily. If you are reading this months from now, re-run the method rather than trusting the numbers, and the ratios will outlive the totals.