A forward deployed engineer (FDE) is a software engineer who works at the customer, not only on the product. An FDE writes code inside the customer's systems, helps sell the solution, deploys it on site, teaches the customer's team to use and extend it, and takes what they learned in the field back into the product. The job is still engineering, but the FDE answers for an outcome at one customer instead of one feature in an app.
The role is easiest to understand as four kinds of work:
- Sales. Technical support for the sales process: working out with the customer what can be built, building a prototype or proof of concept, and answering technical questions before the contract is signed.
- Deployment. Rolling the product out at the customer, often in person. Job postings usually put travel at 25% to 50% of the time.
- Training. Teaching customer teams through workshops and hackathons, so they can keep building after the FDE leaves.
- Back to the product. Turning what they saw in the first three into fixes, features, and reusable patterns that the next customer gets for free.
Below: where the role came from, how it differs from similar jobs, why FDE postings grew faster than almost any other tech title in 2026, what it pays, and when a company actually needs one. Every number has a source at the end.
Where did the forward deployed engineer role come from?
Palantir created the role in the early 2010s. Internally the engineers were called "Deltas", and the official title was Forward Deployed Software Engineer (FDSE). "Forward deployed" is US military jargon for units stationed close to where the action is.
In 2019 Palantir described the difference between a regular software engineer ("Dev") and a Delta in one line: a Dev works on "one capability, many customers", a Delta on "one customer, many capabilities". Deltas sat in Business Development, not in product, and their mandate was a technical outcome at the customer.
According to The Pragmatic Engineer, until around 2016 Palantir employed more FDEs than regular software engineers. When the company launched Foundry, many FDEs moved back into product engineering and brought field experience with them. That customer-to-product loop is still part of the definition.
AI companies picked up the title in 2025. OpenAI set up its FDE team in early 2025, and a16z called the FDE "the hottest job in startups". By 2026 OpenAI, Anthropic, Google, AWS, and Salesforce all hire FDEs, and consulting firms post the most openings.
What does a forward deployed engineer do?
Every company weights the role differently. The Pragmatic Engineer notes that some expect FDEs mainly to close deals, others to make customers successful, others to build the core product. The same four areas still show up across postings.
1. Technical support for sales
FDEs join customer conversations earlier than most engineers. When a prospect is unsure it can use the product, sales offers an FDE to help with the integration. Ramp's posting (quoted by The Pragmatic Engineer) asks FDEs to "collaborate closely with our sales and go-to-market teams to close exciting deals, activate customers, and expand the value Ramp provides over time".
IFS Loops goes further and hires a "Pre-Sales Forward Deployed Engineer" who sits "between presales and product". That person runs discovery, builds proofs of concept "in weeks, not months", presents the same proof of concept to an architect and to a CxO, and helps account executives win technical validation on live deals.
The difference from classic presales: the FDE does not stop at the demo. The proof-of-concept code often becomes the first version of the deployment.
2. Deployment at the customer, often on site
This is the core of the job. Anthropic's London FDE posting asks FDEs to build production applications with Claude "within customer systems" and to deliver MCP servers, sub-agents, and agent skills for production workflows. The same posting: "Travel frequently (25-50%) to customer sites to build in person with customers."
Travel is part of the method. The FDE sits with the people who will use the system, because that is where the real process shows up. OpenAI's head of FDE, Colin Jarvis, told The Pragmatic Engineer that what the customer describes in scoping "often doesn't match the data/system reality on the ground". For a John Deere project, he and an FDE went to Iowa and worked directly with farmers. A former Palantir FDE described working on Airbus's final assembly line.
Palantir expects around 25% of an FDE's time on site. Healthcare AI company Commure estimates up to 50%.
3. Training, workshops, and hackathons
A deployment the customer's team does not understand dies when the FDE leaves. Some postings say so directly. Twilio wants its FDE to "run workshops and hackathons across Twilio's global sales base", to "lead high-energy customer hackathons and builder activation events", and to build reference implementations, starter apps, SDKs, and guides that customer teams can extend on their own. IFS Loops wants its FDE to train presales and run ongoing workshops globally.
AWS describes the customer side of this: many companies already employ capable engineers but need help moving AI projects into production, so AWS FDEs work alongside those teams until the customer can maintain and expand the systems itself (CNBC, October 15, 2026).
4. Field lessons go back into the product
This is what separates an FDE from a consultant or a classic implementation engineer. Anthropic wants FDEs to "identify and codify repeatable deployment patterns and contribute insights back to our Product and Engineering teams". Twilio: "synthesize field insights from every production deployment to directly inform R&D".
The best description of this loop is OpenAI's October 22, 2026 post What FDE means at OpenAI:
- Semiconductors. In mid-2025, about 15 FDEs worked with one semiconductor customer on 10+ chip verification use cases. All of them reached production within 3 to 6 months, and AI-based directed test generation saved about 48% of the time. Once Codex matured, the team moved the use cases onto a common Codex-based harness (plugins and skills instead of separate tools) and removed roughly 90% of the custom code from each project. In one August 2026 case, the first deployment took two months and extending it to a second team took one day.
- Voice customer service. At a large telecom customer, the model scored around 60% on the key evals at the start. FDEs built the platform and defined what the model had to do; Research improved the model. Six months later the score was above 90%, and the resulting voice model now serves other customers too.
- Products from deployments. OpenAI Presence, its product for enterprise voice and chat agents, grew out of customer deployments, through Swarm and the Agents API.
Jarvis's summary: if the FDE team makes itself unnecessary for a class of problems every few months, because the product now handles it, the team is doing its job.
What an FDE engagement looks like
OpenAI splits FDE projects into three phases (per The Pragmatic Engineer):
| Phase | What happens | Output |
|---|---|---|
| 1. Scoping | A few days on site: sitting with users, mapping the process, picking the highest-value areas, prototyping on synthetic data | Priorities and a first prototype |
| 2. Validation | Checking that the scoped problem is really the most valuable one. Evals built with user input and labeling, then hill-climbing on them | A report comparing eval results to the goal, and a go/no-go |
| 3. Delivery | A few days a week on site, building the smallest end-to-end unit that works, demos | A system in production with first users |
Validating before delivery is unusual in classic implementations. Jarvis explains that FDEs work with a lot of ambiguity, so they would rather hit the brick walls early and adjust the scope than spend months building something the customer's data cannot support. Ramp's FDE team turned this into a motto: "always be scoping".
If you are building an AI agent, the validation phase is where evals and traces come in.
Forward deployed engineer vs software engineer, solutions engineer, consultant, and customer success
These roles get mixed up, and job titles blur them further. FDE Notes, which pulls postings straight from employers' applicant tracking systems, counted 919 live roles in customer-facing AI and deployment work on October 25, 2026. Only 204 had "forward deployed" in the title. The largest group (320) was solutions engineering and architecture.
| Role | Main goal | When they are involved | Writes production code in the customer's systems? |
|---|---|---|---|
| Software engineer (product) | One capability for many customers | All the time, on the product team | No, builds the product the customer uses |
| Forward deployed engineer | An outcome at one customer, using many capabilities | From presales through launch and beyond | Yes, that is the core of the job |
| Solutions engineer, sales engineer, solutions architect | Technical advice and demos before the deal closes | Mostly presales | Rarely. At OpenAI, SAs build MVPs and proofs of concept on anonymized or offline data |
| Consultant | Analysis, recommendations, a change plan | Per project | Usually not. Leaves recommendations; an FDE leaves a working system |
| Customer success | Adoption, relationship, renewal | After the sale | No, it is a relationship role, not an engineering one |
A few nuances the table misses:
- FDE vs software engineer. McKinsey's Holger Hürtgen puts it this way: a developer has a predefined product to build, while an FDE, like a consultant, first has to understand the problem. Lightcast estimates the FDE role shares roughly half its skills with software engineering.
- FDE vs solutions engineer. Per The Pragmatic Engineer, FDEs are more hands-on and work with more ambiguity. They write code directly on customer infrastructure using customer tooling, and usually contribute more to the core product.
- FDE vs consultant. Palantir calls this a common mistake: consultants make one-off recommendations, FDEs work with customers long term. Not everyone agrees. a16z's Joe Schmidt writes that at many AI application companies, deployment is led by a professional services employee "sometimes rebranded as a forward deployed engineer".
Why is demand for forward deployed engineers rising now?
The AI deployment gap
OpenAI calls it "the deployment gap": the distance between what models appear capable of and what people can reliably use to do their work. Model capabilities moved faster than the systems and processes needed to use them. Sometimes the missing piece is a product; sometimes a model does well on a benchmark but not with real-world variables and constraints.
An AI agent that works inside a company needs access to its systems, data, permissions, and people. Someone has to sit with the team, connect the agent to the CRM or ERP (for example over MCP), pass the security review, and show people how to use it. Twilio's posting says the FDE should "bridge the gap where many enterprise AI projects stall at the last mile".
Pressure to show a return on AI spending
CNBC reported on October 15, 2026 that after billions spent on infrastructure, investors want business results, and FDEs are the people expected to deliver them. "The Street is starting to become impatient," said Jeff Christian of executive search firm Christian & Timbers. The money followed:
- AWS committed $1 billion to a forward deployed engineering initiative, starting with hundreds of engineers.
- OpenAI launched the OpenAI Deployment Company in May 2026 with more than $4 billion in initial investment, and agreed to acquire Tomoro, bringing roughly 150 FDEs and deployment specialists.
- Anthropic is putting $100 million into Claude Frontier Academy, aiming to train 10,000 FDEs by the end of 2027, mostly engineers already employed by its customers and partners.
- Accenture and Google Cloud are building a 1,000-person FDE workforce, partly by retraining experienced engineers. Accenture's onboarding takes 30 to 90 days.
The job posting numbers
| Source | Data | Period |
|---|---|---|
| Lightcast (via CNBC) | US FDE postings: fewer than 200 in 2023, about 1,200 in 2025, more than 7,300 in the first nine months of 2026 | 2023 to Sept 2026 |
| Indeed (via Business Insider) | 643 postings in April 2025, 5,330 in April 2026, up 729% year over year | Apr 2025 to Apr 2026 |
| Lightcast (via CNBC) | Top posters: Deloitte 872, Accenture 688, Google 499, KPMG 447 | Jan to Sept 2026 |
| FDE Notes | 919 live customer-facing AI roles, 204 with "forward deployed" in the title | Oct 25, 2026 |
Two caveats. These are postings, not hires, and one person can match several listings. And the growth starts from a very small base: FDE is still a thin slice of engineering hiring, and some "new" postings are old implementation roles under a new name. Also notice who posts the most: consulting firms top Lightcast's list, not AI labs.
How much does a forward deployed engineer make?
Lightcast puts the median advertised FDE salary in the US at $195,456 for the first nine months of 2026, up 40% from 2023 and more than three times the median across all US postings. Earlier Lightcast data cited by Fortune compared more than $188,000 for FDEs with roughly $145,000 for traditional software engineers.
| Company and title | Location | Posted range | Notes |
|---|---|---|---|
| Anthropic, Forward Deployed Engineer | London | £225,000 to £255,000 a year | 4+ years of experience, 25–50% travel |
| Accenture, Forward Deployed Engineer | Kraków, Poland | PLN 180,000 to 240,000 a year | Hybrid, posted Oct 19, 2026 |
| Spyrosoft, Forward Deployed Engineer (OpenAI) | Poland, remote | €25.06 to €34.17 net per hour, contractor | Posting has expired |
What skills does an FDE need?
Lightcast groups the requirements in FDE postings into three areas: AI and software (artificial intelligence, Python, software engineering), delivery (workflow management, scalability), and human skills (communication, problem solving, leadership).
What shows up again and again in postings:
- Strong programming. Anthropic asks for Python, ideally plus TypeScript or Java, and experience shipping production applications.
- LLMs in production. Agents, prompt engineering, evaluation frameworks; at Anthropic, also MCP servers, sub-agents, and skills. If those terms are new, start with what an AI agent is and how it works.
- Customer judgment. Caylent CEO Val Henderson (an AWS partner that embeds FDEs with customers) told CNBC the hardest thing to teach is "the customer EQ": understanding the customer, navigating their organization, and making calls when a project goes sideways.
- Experience. Anthropic wants 4+ years in a technical, customer-facing role, or software engineering with consulting or product experience, and explicitly encourages former technical founders. Palantir hires people with as little as a year of work experience, and Ramp hires some exceptional new grads.
- Domain learning. OpenAI says FDEs do not need to know the industry on day one, but it is their responsibility to learn it. On the semiconductor project, new FDEs studied chip design in their spare time.
Does your company need a forward deployed engineer?
If you are reading this as a company trying to put AI to work, not as a candidate, a few questions help.
An FDE makes sense when:
- you bought access to a model or AI platform and months later you still have only a pilot,
- the problem is fuzzy and nobody can write the spec, because only the people doing the work know the process,
- the solution has to run inside your systems (ERP, CRM, internal databases) and pass a security review,
- your team has good engineers but no experience running AI agents in production.
You probably do not need one when an off-the-shelf tool with good documentation does the job, or the problem is well defined and your product team can build it.
There are three usual ways to get this kind of person:
| Option | When it fits | What to watch |
|---|---|---|
| Vendor FDEs (OpenAI, Anthropic, AWS, Salesforce) | A large contract with one vendor | They usually build on that vendor's product, and smaller companies rarely get this level of support |
| A consulting or implementation firm | Large projects that need a team | Check that they leave working code, not just a report, and who maintains it |
| An independent engineer or small team that works this way | Small and mid-size companies, a first process to automate, an MVP | Make sure they actually ship production code and leave knowledge with your team |
Whichever you choose, ask the same questions:
- How do you start: with a workshop and conversations with the people who do the work, or with a ready-made proposal?
- How will we know it works before we build more? Which evals or metrics?
- What stays with us at the end: code, access, documentation, a trained team?
- How often will we see a working version?
- Who maintains the system after launch?
For what a first version costs and how to vet a contractor, see how much an MVP costs in 2026.
When is "FDE" just a new label?
The title's popularity has a downside. Every company has its own flavor of the role, as The Pragmatic Engineer points out. Some FDE postings are classic implementation or no-code configuration work: Lindy, an AI agent platform, expects its FDEs to help customers build no-code workflows "while serving as their technical consultant". Anthropic's Claude Frontier Academy reads FDE as "Frontier Deployed Engineer", which is a training program, not the job.
A simple test: does this person write code that runs in production at the customer, and do they own the result after launch? If yes, it is FDE work, whatever the title says. If the work ends at a demo or a recommendation, it is presales or consulting.
CNBC also notes that hard evidence of financial returns from FDE projects is still thin. Case studies show productivity gains, but they are reported by the companies themselves.
Frequently asked questions
What does FDE stand for?
FDE stands for forward deployed engineer, an engineer stationed with the customer. Palantir, which created the role, uses the title Forward Deployed Software Engineer (FDSE) and internally called these engineers Deltas.
What is the difference between a forward deployed engineer and a software engineer?
A product software engineer builds one capability for many customers. An FDE works for one customer and uses many capabilities of the product to reach a specific outcome. An FDE also helps with sales, deploys on site, trains the customer's team, and feeds lessons back into the product.
Is a forward deployed engineer the same as a solutions engineer?
No. A solutions engineer or solutions architect works mostly before the deal closes: advising, running demos, building proofs of concept, and rarely writing code on the customer's infrastructure. An FDE also supports sales, but the main job is building and deploying a production system in the customer's environment.
How much does a forward deployed engineer make?
Lightcast puts the median advertised US FDE salary in 2026 at $195,456 a year. Anthropic's London FDE posting lists £225,000 to £255,000. Postings in Poland in 2026 range from PLN 180,000 to 240,000 a year at Accenture in Kraków to about €25 to €34 net per hour for contractors at Spyrosoft.
Do forward deployed engineers travel a lot?
Usually. Postings range from about 25% of the time (Palantir) to 50% (Commure). Anthropic's London posting says 25% to 50% travel to customer sites.
How do you become a forward deployed engineer?
The common path is several years of software engineering, ideally with LLMs in production, plus direct customer work (consulting, implementations, or your own startup). Anthropic explicitly encourages former technical founders to apply. Accenture and Google retrain experienced engineers in 30 to 90 days.
How this looks in my own work
I am not an FDE at a large company, but I work with clients in a very similar way. We start with a workshop where we map the process with the people who do it. Then we write down the scope, and I work in two-week sprints on real data, showing a working version after each sprint. I design the UX and write the code myself: AI agents, MVPs, and web and mobile apps. I work from Kraków, Poland, as a B2B contractor.
If you have a process you want to hand to an agent, or AI to add to an existing product, see AI agent development and AI integration, or just tell me what you want to build.
Sources
Accessed October 26, 2026 unless noted.
- OpenAI, Colin Jarvis, What FDE means at OpenAI (Oct 22, 2026)
- The Pragmatic Engineer, Gergely Orosz, What are Forward Deployed Engineers, and why are they so in demand? (Aug 12, 2025)
- Palantir Blog, Dev versus Delta: Demystifying engineering roles at Palantir (2019)
- a16z, Joe Schmidt, Trading Margin for Moat: Why the Forward Deployed Engineer Is the Hottest Job in Startups (2025)
- CNBC, AWS, Anthropic race to deploy army of specialized engineers (Oct 15, 2026), Lightcast data
- Lightcast, Q: What is a Forward Deployed Engineer? A: The fastest growing AI job
- Business Insider, Job postings for this tech job have grown over 700% in the last year (May 2026), Indeed data
- Fortune, One of the fastest-growing jobs in Silicon Valley sends engineers straight to customers' offices (Sept 3, 2026)
- FDE Notes, Forward-deployed engineering market data (snapshot Oct 25, 2026)
- McKinsey, What is a forward-deployed engineer?
- Anthropic, Forward Deployed Engineer, London, job posting
- Twilio, Forward Deployed Engineer, job posting (Built In)
- IFS, Pre-Sales Forward Deployed Engineer, IFS Loops, job posting
- Accenture, Forward Deployed Engineer, Kraków, job posting (posted Oct 19, 2026, via posada24)
- Spyrosoft, Forward Deployed Engineer (OpenAI), job posting (justjoin.it, expired)
Want to build an AI agent, an MVP, or an app?
I design and build AI products from the first version to production: MVPs, React Native mobile apps, web apps, and AI agents. Tell me what you want to build and I will tell you where to start.