The most sought-after engineers in AI don't train models. They go and sit inside a customer's business until a model does something useful there. In 2026, OpenAI, Anthropic, AWS and Microsoft each announced a business built around that work. The companies with the most capable models have concluded that selling access isn't enough. Whether you want the job or you're trying to get AI working in your own organisation, that conclusion matters.
What is a forward deployed engineer?
A forward deployed engineer (FDE) is a software engineer, employed by a technology company, who works inside a customer's organisation, alongside the people who do the work, building the company's product into the customer's real systems until it runs in production and changes how the work gets done.
OpenAI's job posting puts it plainly. Its FDEs "lead complex end-to-end deployments of frontier models in production alongside our most strategic customers", and they own "discovery, technical scoping, system design, build, and production rollout". The measure isn't a demo or a signed contract. It's "production adoption, measurable workflow impact, and eval-driven feedback that changes product and model roadmaps."
This guide is for two readers. If you want the job, or want to understand it, the first half covers the work, the skills, the pay and the way in, from the companies' own job postings. If you lead an organisation and keep hearing the term, the second half covers what the role tells you about deploying AI, and how to get the same capability without hiring one.
What does "forward deployed" mean?
The phrase comes from US military jargon: the US military keeps forward deployed forces stationed abroad. Palantir, which made the job title famous, took it literally: its 2020 prospectus says its forward deployed engineers "have travelled to bases in Afghanistan and factories in the industrial Midwest to deploy our platforms." Palantir still calls the team Delta.
Where did the forward deployed engineer role come from?
Palantir, the data software company. Its current posting for the role says: "We pioneered this unique position, embedding talented engineers directly with our customers to tackle their most pressing challenges head-on."
Accounts differ on the year. Andreessen Horowitz says that in 2011 Palantir took its solutions engineers and integration engineers, "typically among the lower-status roles in any engineering org", and gave them the new title. Wikipedia cites a 2009 article that already used it.
The prospectus explains what Palantir got in return: "Time in the field adds to the continuous improvement of our platforms. As FDEs help customers make the most of our software, they observe users' challenges firsthand." That two-way trade is the whole model. The customer gets software that works in its world, and the vendor learns what to build next. It's why AI companies copied it.
Palantir still runs on it. On 8 October 2026, 79 of the 313 jobs on its job board had "forward deployed" in the title. It pairs FDEs with a non-engineering role, the deployment strategist, who goes onsite to meet customer analysts and works "with Forward Deployed Engineers to integrate the data into a stable and extensible pipeline". If you've searched for "AI deployment strategist", that's the template.
Why are AI labs hiring forward deployed engineers now?
A frontier model is general. A customer's work is specific: its own data, systems, controls and habits. Somebody has to close that gap, and in 2026 the AI companies decided to staff it themselves.
- OpenAI. Its Forward Deployed Engineering team "partners with customers to turn research breakthroughs into production systems." On 11 May 2026 it launched the OpenAI Deployment Company, with more than $4 billion of initial investment, to "extend OpenAI's ability to embed engineers specialized in frontier AI deployment, known as Forward Deployed Engineers". It also agreed to buy Tomoro, an applied AI consulting and engineering firm, which brings "approximately 150 experienced Forward Deployed Engineers and Deployment Specialists". The announcement is blunt: "building powerful AI models is only part of the work."
- Anthropic. Its FDEs embed "directly with our most strategic customers to drive transformational AI adoption", and its September 2026 London posting describes them as "founding FDEs" shaping the practice. On 4 May 2026 it announced a separate AI services company with Blackstone, Hellman & Friedman and Goldman Sachs to bring Claude to mid-sized companies, because "putting Claude to work in an organization's core operations takes hands-on engineering and deep familiarity with how each business runs."
- The cloud providers. According to CNBC, AWS is putting $1 billion into a new forward deployed engineering unit seeded with "thousands" of FDEs, and Microsoft is committing $2.5 billion and 6,000 employees to a new unit, Microsoft Frontier Co., whose staff will be embedded with clients. Google Cloud's chief executive has said it is hiring additional FDEs, and that demand from customers and partners for its engineers' help "is growing very rapidly", as reported by Fast Company.
The companies give two reasons in their own words. The first is adoption. OpenAI says more than one million businesses have adopted its products and APIs, and that "the next stage of enterprise AI will be defined by how effectively businesses can deploy this technology into real-world use cases." The second is learning. OpenAI asks its FDEs to "share field feedback that helps Research and Product understand where the models succeed and where they can improve." Anthropic asks its FDEs to "identify and codify repeatable deployment patterns" for its product and engineering teams.
Some of the boom is relabelling. Andreessen Horowitz calls Palantir's move "title arbitrage": a new title that signals a role now matters more. SVPG, a product-management firm, adds a warning about Palantir's own model: if all it had were FDEs, "they would quickly end up with thousands of large, bespoke solutions that would need to be maintained and improved indefinitely." The role works when what FDEs learn flows back into one product.
What does a forward deployed engineer do day to day?
The postings agree on the shape of the work, if not every detail.
- Discovery. Sitting with the people who do the work and finding where AI would make a measurable difference. Anthropic asks for "strong communication skills to conduct discovery with customers".
- Building in the customer's systems. This is production engineering, not prototyping for show. Anthropic's FDEs "work within customer systems to build production applications with Claude models" and deliver "MCP servers, sub-agents, and agent skills that will be used in production workflows." Palantir's spend their time "wrangling massive-scale data" and "developing custom applications tailored to customer needs".
- Getting it used. OpenAI's FDEs "embed closely with customer teams, understand their needs, and guide adoption of what you build".
- Feeding it back. Turning what worked into "tools, playbooks, or building blocks that others can use", and telling product and research teams where the models fall short.
- Travel. Palantir asks for up to 25%, Anthropic 25 to 50%, and OpenAI up to 50%.
Engagements follow a pattern. OpenAI says a typical Deployment Company engagement begins with "a focused diagnostic of where AI can create the most value", narrows to "a small number of priority workflows", and then puts FDEs "inside the organization to design, build, test, and deploy production systems, connecting OpenAI models to the customer's data, tools, controls, and business processes". AWS, according to CNBC, embeds pods of roughly five or six engineers in a customer at a time, and aims to leave behind self-sufficient teams within weeks.
What skills does a forward deployed engineer need?
The postings ask for three things.
Engineering that ships. Every posting asks for production code. OpenAI wants FDEs who "write and review production-grade code across frontend and backend using Python, JavaScript, or comparable stacks". Anthropic wants "proficiency in Python" and "experience shipping production applications". In Bloomberry's analysis of 1,000 FDE job postings, published in November 2025, Python appeared in 66%, TypeScript in 35% and AWS in 32%.
Production experience with AI models. Anthropic asks for "production experience with LLMs including advanced prompt engineering, agent development, evaluation frameworks, and deployment at scale". OpenAI wants people who have "built or deployed systems powered by LLMs or generative models and understand how model behaviour affects product experience".
Working with customers. That means discovery, explaining trade-offs in plain terms, and judgement under pressure. OpenAI asks FDEs to "make trade-offs between scope, speed, and quality". Anthropic asks for "high agency with an ability to navigate ambiguity present in complex organizations".
The experience bar varies by company. Palantir asks for "1+ years of relevant, post-college work experience" and hires new graduates and interns into the role. Anthropic asks for 4+ years in a technical, customer-facing role. OpenAI asks for 5+ years of engineering or technical deployment experience that includes customer-facing work.
Forward deployed engineer vs solutions engineer, sales engineer and consultant
The roles overlap, and the AI labs often put several of them on one account. OpenAI's FDEs "collaborate closely with Sales, Solutions Engineering, Solutions Architects, and Customer Success Managers who work on the same account." The clearest differences are when each one works and what they're measured on.
| Solutions or sales engineer | Forward deployed engineer | Consultant or systems integrator | |
|---|---|---|---|
| When | Before the sale | After the sale, through to production | For an agreed scope |
| Works for | The vendor | The vendor | The client |
| Where | Mostly at the vendor, with occasional customer visits | Inside the customer's systems, often 25 to 50% travel | At the client |
| Builds | Demos and proofs of concept | Production systems on the vendor's product | What the client's specification asks for |
| Measured on | Closing the deal | Production use and workflow impact | Delivering the agreed scope |
At OpenAI, a solutions engineer is "the technical expert in the sales process", working with the account team to accelerate "deal closure". Anthropic's pre-sales equivalent, the Applied AI Architect, builds "prototypes and proof-of-concepts" and travels "occasionally to customer sites". SVPG describes the classic custom-build firm this way: "The client provides the spec, they agree on a price", and the firm "builds what's requested".
The line is blurrier than the table suggests. Pay overlaps: Anthropic advertises its US pre-sales architect role at $240,000 to $315,000. And if Andreessen Horowitz is right, the first FDEs were solutions and integration engineers with a new title. The real difference is commitment: an FDE is on the hook until the software is in production and people use it.
How much does a forward deployed engineer make?
Pay depends on the company, the level and the city. These are the advertised base salary ranges on the companies' own job boards when we checked on 8 October 2026. Equity, bonuses and sign-on payments come on top where the company offers them.
| Company and role | Location | Base salary range |
|---|---|---|
| Palantir, Forward Deployed Software Engineer | New York | $135,000 to $200,000 (Palantir's estimate) |
| OpenAI, Forward Deployed Engineer | San Francisco | $185,000 to $300,000, plus equity |
| Anthropic, Forward Deployed Engineer | London | £225,000 to £255,000 |
| OpenAI, Manager, Forward Deployed Engineering | New York | $302,000 to $385,000, plus equity |
| Anthropic, Manager, Forward Deployed Engineering | New York | $320,000 to $405,000 |
OpenAI's sector FDE roles (healthcare, legal and financial services) in New York, San Francisco and Seattle list the same $185,000 to $300,000 range. Across the wider market, Bloomberry found a median of $173,816 among the FDE postings that disclosed a salary.
How do you become a forward deployed engineer?
None of the postings we read asks for a certification. They ask for evidence that you've shipped software and that customers trust you with it.
- Ship production software first. The labs want years of it: 4+ at Anthropic, 5+ for OpenAI's FDE role and 7+ for its forward deployed software engineer role.
- Build with models, end to end. Prompts, agents, evaluations and deployment, on real data. Anthropic's postings name MCP servers, sub-agents and agent skills as things its FDEs deliver.
- Get customer-facing experience. Solutions engineering, consulting or a startup's first customers all count. Anthropic invites "former technical founders" to apply, and OpenAI lists being a "former founder, or early engineer at a startup who has built a product from scratch" as a plus.
- Learn an industry. Anthropic lists a background in financial services or healthcare and life sciences as a plus, and OpenAI recruits FDEs by sector: healthcare, legal and financial services.
- Start where the door is open. Palantir hires new graduates and interns as forward deployed software engineers. That's the clearest entry route. The AI labs mostly hire experienced engineers.
- Be honest about the trade-offs. Expect a quarter to half of your time travelling, and customer deadlines rather than your own roadmap. Wikipedia notes that some engineers see the role as undesirable for exactly those reasons.
What the forward deployed engineer tells leaders about AI deployment
Look at what the labs did. They have the most capable models and their own researchers on hand, and they still concluded that the model isn't the hard part. OpenAI: "building powerful AI models is only part of the work." Anthropic: putting Claude to work in core operations "takes hands-on engineering and deep familiarity with how each business runs."
Research on enterprise AI points the same way. MIT NANDA's 2025 study drew on more than 300 public AI initiatives, interviews at 52 organisations and a survey of 153 senior leaders, and its authors call the findings preliminary. It found that 60% of organisations had evaluated enterprise-grade AI systems, 20% reached a pilot, and 5% reached production. Most failed "due to brittle workflows, lack of contextual learning, and misalignment with day-to-day operations." The divide, the authors wrote, "does not seem to be driven by model quality or regulation, but seems to be determined by approach."
That's the job an FDE does. It isn't about a better model. It's someone at the desk where the work happens: fitting the tool to the workflow (and often the workflow to the tool), connecting the data, and staying until people use it.
The catch is who gets one. OpenAI and Anthropic both say their FDEs work with their "most strategic customers". Anthropic set up a separate company for mid-sized businesses because firms "from community banks to mid-sized manufacturers and regional health systems" can gain from AI "but lack the in-house resources to build and run frontier deployments." If you run a 200-person company, a lab engineer probably isn't coming to sit with your team. You can still build the capability.
How to get forward deployed capability without hiring an FDE
What makes an FDE effective isn't the title. It's five conditions: someone who understands the work, someone who can build, time spent side by side, a measure based on production use, and an owner once the builder leaves. You can create all five. There are six steps, in order.
01 · Pick one workflow, not one tool
FDE engagements start with a diagnostic and narrow to a few priority workflows. Do the same. Trace the work back from the customer and find the one step that limits everything else. Tracing work back from the customer, in module 4 of our free course, shows how, and the module ends with one bottleneck and a value put on it.
02 · Put a builder next to the people who do the work
The FDE's advantage is proximity. Pair someone who can build (an engineer, an analyst who scripts, or an outside builder) with the people who run the workflow, and protect their hours together. Your power users are often the best place to start: What your power users are telling you shows how to find them and what to do with what they've built.
03 · Redesign the work before you automate it
OpenAI says its Deployment Company's FDEs will "redesign organizational infrastructure and critical workflows" around AI. They don't just bolt a model onto what exists. Automating a broken step makes the mistakes faster. Redesign before you automate shows how to fix the work first.
04 · Judge it on production use, not the demo
OpenAI measures its FDEs on production adoption and workflow impact. Hold your own build to the same test: first a sandbox that can't touch live systems, then a limited live trial, kept or killed on the evidence. That's the two-stage proof in module 6 of our course.
05 · Name the owner before the builder leaves
FDEs move on to the next customer. AWS says its aim is to leave behind self-sufficient teams. Give everything you build an owner who answers for it and a backup who can stop it: Owners and backups.
06 · Borrow the builder if you don't have one
If nobody in-house can build, bring someone in for the build and keep the ownership. In MIT NANDA's sample, external partnerships reached deployment about 67% of the time, against about 33% for internal builds. The authors warn that these outcomes are self-reported and may reflect the organisations that chose each route as much as the route itself.
That last step is the gap we work in. Deployment Labs is a team of AI deployment engineers. We onboard teams to OpenAI, Claude and Codex, sit with the people who do the work, build the workflows, skills and plugins they need, and hand them to an owner who keeps them running. See how we work, or read our guide to AI deployment for the whole method.
What's still open
- Whether the role lasts. The labs want FDEs to turn field work into reusable patterns and product features. If that works, each deployment needs less custom engineering and the role narrows. If it doesn't, the FDE businesses start to look like consulting firms, which is the risk SVPG describes.
- The Tomoro acquisition. OpenAI said in May 2026 that the deal was "expected to close in the coming months", subject to regulatory approvals. We haven't seen confirmation that it has closed.
- Pay. Ranges are moving with demand. The figures here are from postings as listed on 8 October 2026.
We'll update this article when pay ranges shift materially or the labs' deployment companies report on their work. The date at the top of the page shows when we last checked.
Sources
- Forward Deployed Software Engineer, New York, Palantir, October 2026: the role description, requirements, travel and estimated salary range.
- Deployment Strategist, New York, Palantir, October 2026: the deployment strategist role, its work with FDEs, travel and salary range.
- Form S-1 registration statement, Palantir Technologies, August 2020: FDEs at bases in Afghanistan and factories in the Midwest, and what field time does for the product.
- Forward Deployed Engineer (FDE), San Francisco, OpenAI, September 2026: responsibilities, how success is measured, requirements, travel and salary range.
- Forward Deployed Software Engineer, San Francisco, OpenAI, November 2025: the roles that share an account, and the experience asked for.
- Manager, Forward Deployed Engineering, New York, OpenAI, September 2026: manager salary range.
- Partner AI Deployment Engineer, AWS, OpenAI, June 2026: what OpenAI's AI Deployment Engineering team does.
- Solutions Engineer, Ads Solutions, San Francisco, OpenAI, September 2026: the solutions engineer's role in the sales process.
- OpenAI launches DeployCo to help businesses build around intelligence, OpenAI, May 2026: the OpenAI Deployment Company, its investment, the Tomoro acquisition and the shape of an engagement.
- Forward Deployed Engineer, London, Anthropic, September 2026: responsibilities, requirements, travel and salary range.
- Manager, Forward Deployed Engineering, New York, Anthropic, 2026: manager salary range.
- Applied AI Architect, Commercial, Anthropic, August 2026: the pre-sales role and its salary range.
- Building a new enterprise AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs, Anthropic, May 2026: why deployment takes hands-on engineering, and the gap for mid-sized companies.
- The GenAI Divide: State of AI in Business 2025, MIT NANDA, July 2025: evaluation, pilot and production rates, the reasons for failure, and partnership against internal build outcomes (copy hosted by AI News).
- Forward-deployed job titles, Andreessen Horowitz, January 2026: the 2011 retitling at Palantir and "title arbitrage" (secondary).
- Forward Deployed Engineers, SVPG, September 2025: FDEs and the risk of bespoke solutions (secondary).
- AWS puts $1 billion into new AI unit to embed engineers with customers, CNBC, June 2026: AWS's forward deployed engineering unit (secondary).
- Microsoft commits $2.5 billion and 6,000 employees to new AI implementation unit, CNBC, July 2026: Microsoft Frontier Co. and the military origin of "forward deployed" (secondary).
- Google, Box CEOs say this is the "most in-demand" job in tech, Fast Company, May 2026: Google Cloud's FDE hiring (secondary).
- I analyzed 1,000 forward deployed engineer jobs, Bloomberry, November 2025 (updated January 2026): skills in FDE postings and the median advertised salary (secondary).
- Forward deployed engineer, Wikipedia, October 2026: the term's military origin, the 2009 usage and engineers' reservations about the role (secondary).