If you have spent any time inside an AI rollout in the last twenty-four months, you have watched the same movie. The vendor demo blows the room away. Procurement signs. Six months later, the pilot is still a pilot. Nobody is using it. Nobody can say why.
That gap, between a working demo and a system that actually runs inside a real company, is the most expensive problem in software right now. The role built to close it is the Forward Deployed Engineer.
What is a Forward Deployed Engineer?
A Forward Deployed Engineer (FDE) is an embedded technical professional who works inside a customer's environment to take an AI or software product from discovery to running, measurable production. They are half engineer, half consultant, half product manager. They do the work that ships.
The job is end to end. An FDE runs discovery interviews with the customer, designs the architecture, integrates with the customer's existing stack, builds prototypes, deploys to production, measures impact, and feeds what they learned back to the product team. They do not stop when the contract closes. They stop when the system is actually working and the metric is moving.
Where the role came from
Palantir invented the modern shape of this job. Inside Palantir, the function was called Forward Deployed Engineer (or, in the early days, a Delta). The Deltas were the engineers Palantir sent into the customer (an intelligence agency, a bank, a hospital network) to make the platform actually do something useful in that specific environment. They were not pre-sales. They were not implementation. They were engineers who lived at the customer and shipped.
For about a decade, FDE was a Palantir-flavored term. It mostly stayed there.
Then generative AI happened, and every company on earth tried to buy it at once. By 2024, it was clear the same gap Palantir had been solving in defense and finance was now showing up at every Fortune 1000 buying ChatGPT, Claude, or a Copilot, and at every Series B startup trying to ship an AI product. The models were good. The deployments were not. Somebody had to live inside the customer and make it work.
That somebody is the FDE. The role moved from a niche Palantir job title to a category. By 2025 it was a hiring priority at OpenAI, Anthropic, Salesforce, Google, ServiceNow, and a long tail of AI-native startups. The Palantir mold became the industry standard.
What a Forward Deployed Engineer actually does (a week in the life)
Job descriptions for the role read like four different jobs taped together. That is because the work really does span four different kinds of days.
Monday. Discovery. You are on calls with the customer's operations team, the line-of-business owner, and probably one skeptical SVP. You are not pitching. You are mapping their workflow, identifying where AI can actually move a metric, and ranking opportunities by impact versus effort. You leave with a one-page workflow map you will refine all week.
Tuesday. Integration work. The customer has Salesforce, a homegrown order system, a data warehouse, and an internal documentation tool nobody has cleaned up since 2019. You are writing the glue. Webhooks, queues, retries, auth, idempotency. The unsexy plumbing that decides whether the AI feature ever sees real data.
Wednesday. Building. You are writing a RAG pipeline, wiring up tool calls, or shaping an agent that needs to do three things in sequence without hallucinating. You are also writing evals, because nothing about this goes to production without a way to measure regression.
Thursday. Stakeholders. You are running a steering committee. You are translating "the retrieval recall dropped four points after we changed chunking" into "the assistant gets the right answer eight percent less often, and here is the fix." You are writing a design doc that the customer's CTO will read at 11pm.
Friday. Deploy and observe. You are pushing the system into a real environment with structured logs, tracing, metrics, and a rollback plan. You are watching what real users do with it. You are quietly screenshotting the moments where the customer's team finally trusts the system, because those screenshots become the case study that lands the next deal.
Forward Deployed Engineer vs. Solutions Engineer vs. Implementation Manager vs. Consultant
This is the comparison most people are searching for, because the titles get muddied.
| Role | Where they work | Focus | Deliverable | When they stop |
|---|---|---|---|---|
| Forward Deployed Engineer | At the customer, end to end | Bring AI to real production | Running system and a moved metric | When impact is measurable |
| Solutions Engineer | Pre-sales, demos, onboarding | Show feasibility, win the deal | Demo, POC, initial integration | When the contract closes |
| Implementation Manager | Post-sale, project management | Coordinate rollout | Project plan, status, handoff | When the project is "live" |
| Consultant | At the customer, high level | Diagnosis and recommendation | Slide deck, strategy doc | On report delivery |
| Product Engineer | Inside the product | Generic features for everyone | Feature in the product | When it hits GA |
The Solutions Engineer wins the deal. The Implementation Manager runs the schedule. The Consultant writes the recommendation. The FDE ships the working system and stays accountable for the result.
FDEs own outcomes, not artifacts.
FDEs own outcomes, not artifacts.
Why the role exploded in 2024-2026
Job postings up 4,200% since 2023. Roles tagged Forward Deployed Engineer grew roughly forty-two times their 2023 baseline. In 2025 alone, postings jumped another 1,165%. That is not a niche role. That is a category being institutionalized in real time.
The big labs are all in. OpenAI lists around thirty Forward Deployed-related roles at any given time and acquired Tomoro specifically for FDE talent in 2026. Anthropic posts Forward Deployed Engineer, Applied AI roles embedded in strategic customers as a stated go-to-market motion. Google announced plans to hire hundreds of FDEs in May 2026 to help customers adopt its AI products.
The application-layer companies are matching them. Salesforce committed publicly to building a team of 1,000 Forward Deployed Engineers. ServiceNow and Accenture launched a joint FDE program to scale agentic AI across the enterprise. Palantir, the originator, continues to expand the function.
The reason underneath all of it: 70% of AI pilots never reach production. That number has been repeated by analysts, McKinsey, and basically every CIO conference in 2025. Companies are buying AI faster than they can deploy it. The Forward Deployed Engineer is the role designed to fix that ratio.
When seventy percent of pilots fail, every successful deployment is worth a fortune. Whoever can land them becomes the most valuable hire on the AI go-to-market team.
Who hires Forward Deployed Engineers and what they pay
Frontier AI labs. OpenAI, Anthropic, Mistral, Cohere. These are the highest-comp roles. An FDE at OpenAI in San Francisco or at Anthropic Applied AI sits inside strategic customers and turns model capability into real production systems. Total comp for senior FDEs at the labs typically lands in the range of $300K to $600K base plus equity, with the top of the band reserved for people who can be the technical lead on a Fortune 500 deployment from day one.
Enterprise AI platforms. Palantir, Salesforce (Agentforce), ServiceNow, Databricks, Snowflake. Comp here typically runs $180K to $350K base plus equity and a variable component. Palantir's original role still exists and still pays well. Salesforce is scaling fastest.
Hyperscalers and consultancies. Google Cloud, AWS, Microsoft, Accenture, Deloitte, McKinsey QuantumBlack. Comp ranges from $150K to $300K depending on level. These roles often involve more travel and a heavier portfolio of customers per FDE.
AI-native startups. A long and growing list. Often Series B and later. Comp varies wildly. Equity is usually the real upside.
Numbers move. Treat the ranges as directional, not gospel. FDE comp now sits consistently at or above senior software engineer comp, often well above, because the role compounds revenue impact in a way pure product engineering does not.
The skills a Forward Deployed Engineer actually needs
The role does not ask for a unicorn. It asks for three coherent capability stacks that most people only have one or two of. The job is to build all three.
Stack one: technical depth
You do not have to be the researcher who fine-tunes the model. You do have to ship systems that use models, in production, with real data. That means:
- AI system architecture for non-researchers. RAG, agents, function calling, evals, guardrails.
- APIs, webhooks, queues, retries, idempotency. The plumbing.
- Enough SQL and data work to debug the pipeline yourself when it breaks.
- Deploy and observability. Containers, secrets, structured logs, tracing, rollback.
- Security and compliance baseline. Enough to not get blocked by the customer's IT review.
These map directly to the FDE School curriculum, which is organized around exactly this skill set.
Stack two: customer presence
This is the half that engineers traditionally hate and that determines whether you actually succeed in the role.
- Technical discovery. Sitting with the customer, mapping the workflow, finding the real friction point instead of the one they think they have.
- Stakeholder management. Working with operators, IT, security, and the C-suite in the same week.
- Executive communication. Writing the design doc the CTO will read, and the slide the CFO will sign off on.
- Change management. Knowing that adoption is not a launch event, it is a six-month grind.
FDE is the role where customer presence stops being a soft skill and starts being a deliverable.
Stack three: systems thinking
The third stack is what separates a working FDE from a brilliant one.
- Problem decomposition under ambiguity. The customer's brief is always wrong. Your job is to make it right.
- Solution design and trade-off analysis. Cost, latency, accuracy, maintainability. You will trade three of those four every week.
- Doing things that don't scale, systematically. You will hand-curate the first eval set, the first prompt library, the first ten test cases. You will then build the system that makes it scale to ten thousand.
- Feedback loops. Representing the customer back to the product team so the platform gets better, not just this one deployment.
An FDE who masters all three stacks becomes irreplaceable.
How to become a Forward Deployed Engineer
The role attracts four kinds of people. Each one has a real path in. None of them are wrong.
AI is pushing every knowledge worker toward forward-deployed work. The consultants, PMs, and operators retooling into this role now are the ones who will own the next decade.
From software engineering
You already have stack one. What you are missing is customer presence and the consulting craft. The fastest path is to volunteer for any customer-facing engagement your current company will let you near. Solutions Engineering rotations. Customer Engineering. Field engineering. Internal tools work that requires you to interview the people who will use the thing.
The trap to avoid: assuming the technical bar is the whole job. The technical bar is the table-stakes half. FDE hiring panels filter on the customer presence half. Practice it on purpose.
From product management
You already have customer presence and systems thinking. You are probably under-leveled on the technical depth. The path in is to actually build, end to end. Not just spec it. Build a real RAG pipeline. Ship an agent with evals. Deploy something to a real server with real observability. Once you have an artifact you can show, you are competitive for the role at most labs.
From pre-sales or Solutions Engineering
This is the shortest path. You already do discovery, demos, and stakeholder management. You probably write some code, just not full systems. Lean hard into the engineering half. Build a portfolio that shows you can ship to production without an engineer babysitting you. The hiring managers know your title is one rung below where the role is going. They are looking for the proof you closed the gap on your own.
From consulting
The hardest pivot and the most valuable when it lands. You already do problem decomposition, executive communication, and ambiguity for a living. You probably do not ship code. The path is the same as the PM path, but you have to fight harder against the perception that you are a "talker." A real shipped portfolio is the only thing that breaks that perception. Build it. Show it.
Whichever path you are on, the structure is the same. Identify which of the three stacks you are weakest in. Build evidence in that stack you can point at. Apply for roles that explicitly want a hybrid profile (the labs almost all do).
Is Forward Deployed Engineer the right role for you?
You will thrive in the role if:
- You like ambiguity. The customer's problem statement is always wrong on day one, and you find that interesting instead of frustrating.
- You want to see the result of what you build, in production, used by real people, moving a real metric. Building features that go into a roadmap and ship eight quarters later is your nightmare.
- You are energized by the customer relationship, not drained by it. You enjoy the room, the executive call, the steering committee.
- You are comfortable being the most technical person in the business conversation and the most business-fluent person in the technical conversation.
- You want compounding career equity. FDEs build a Rolodex of senior buyers and operators who will hire you, fund you, or partner with you for the next twenty years.
You will burn out in it if:
- You want deep focus time and a quiet IDE. The role is fragmented by design.
- You hate travel, customer dinners, and stakeholder calls. You will do all three.
- You need a tightly defined scope. The role is the opposite of that.
- You want pure technical depth and no business context. Go be a research engineer or a platform engineer. Both are great jobs. They are not this one.
This is one of the highest-leverage roles in AI right now, and the highest-context-switching one. Both are true at the same time.
Where to go next
The companies setting the AI agenda are hiring for it now, in volume, at premium compensation. The skill stack is teachable. The professional network does not yet exist in any organized form.
That is the gap FDE School fills. The course trains the three stacks. The certification gives you an artifact employers recognize. The network is three layers deep: your cohort retooling alongside you, the alumni already placed at frontier AI labs and the companies hiring this role, and the working FDEs at Palantir, OpenAI, Anthropic, Salesforce, Google, and ServiceNow who teach the program and open doors inside their own teams. Hiring for this role is a network move. That is the one we built.
If you are already in the role, FDE School is the professional home you have been missing. If you are trying to break in, it is the most direct path on the market.
Join the waitlist for early access to the program, the community, and the editorial drops. No spam. Technical drops every two weeks. Unsubscribe anytime.