What forward deployed engineering is, and how it works
Forward deployed engineering (FDE) is a way of delivering software in which engineers work inside a client’s operation, learn how the work is really done, and build production systems with the client’s own data and people. The engagement ends when the client’s team can run the system without the engineers.
Written by the Verne engineering team.
Last reviewed
What a forward deployed engineer does in the first week
The work starts on the floor, not in a requirements document. By the end of the first week the team has a measured process, one agreed number and a first fix.
Meet the people who do the work
No laptop and no architecture talk in the first two days. The people on the floor decide how much the engineers get to see.
Shadow the job
Sit beside each role at a quiet hour and a busy one, and write down every workaround and side spreadsheet.
Time every step twice
What the procedure says, and what a stopwatch shows. A step at twice its procedure time is the first suspect.
Agree one number
Success means one measure falls from today’s baseline to a target, read from a log the business already trusts.
Ship a first fix
One visible change that removes the most wasted minutes, live within the first week or two.
Three roles, one stuck order
At 6 pm an order is stuck at the gate and the trucks are waiting. Play the next four weeks and watch who moves it, and what each role leaves behind.
Core engineerAt head office
Order waitingTicket filed. The gate clerk reports an error on the release screen. It joins the support queue.
ConsultantIn meetings
Order waitingNothing has happened yet.
Forward deployed engineerOn the floor
Order movingOrder released. The trucks leave the same evening.
+1 hour. Core engineer: Ticket filed, order waiting. Consultant: not started, order waiting. Forward deployed engineer: Order released, order moving.
How it differs from consulting, outsourcing and extra staff
A forward deployed engineer works one customer across many capabilities. A product engineer works one capability across many customers. The difference shows in what each model leaves behind.
| Forward deployed team | Consultant | Staff augmentation | Outsourced project | |
|---|---|---|---|---|
| What they produce | A working system on your data, and a team that can run it | Analysis, a recommendation and a plan | Code on the tasks you set | Software that meets the specification |
| Code and IP at the end | Handed to you at full delivery | Usually no code | Yours, as the work is done | Depends on the contract |
| How success is measured | One agreed metric, daily use and a handover test | Acceptance of the recommendation | Tickets closed | Sign-off against the specification |
Compare six models, including solutions engineers and in-house hires
Four phases, then run and improve
Discover
Learn how the work is really done, measure it, and agree the metric and the scope boundary.
Deploy
Named accounts with scoped access, a decision boundary in writing, and a choice of cloud, on-premises or hybrid.
Build
One working loop on real data, with the test set built first, rules deciding, and a person approving what matters.
Transfer
A shadow partner from week one, a one-page runbook, a fire drill, and two weeks of your team running it alone.
Run and improve
Hyper-care after go-live, then adoption and health watched so the system stays in daily use.
When it fits, and when it does not
Embedding a team is worth its cost when the work is messy in ways a specification cannot capture. Answer six questions about one workflow.
1 Does the work run through many exceptions nobody wrote down?
Partial deliveries, overrides, missing documents, special customers.
2 Does the data sit in several systems that should agree?
An ERP, a CRM, email, a portal and a spreadsheet, for example.
3 Do staff keep side spreadsheets or notes to get the job done?
Each one marks a gap in the official system.
4 Could success be measured as one number?
Hours a week, days to close, errors a month.
5 Will someone on your side own the system after launch?
A person, or a team, who will run and change it.
6 Does a product you could buy already do this job well?
Not a product that needs heavy custom work to fit.
Answer the six questions about one workflow.
What it costs to try
Every Verne engagement starts with a two-week diagnostic, a fixed fee of US$2,000–5,000. It ends with a measured cost, a designed fix and a fixed price for the build, in writing. A forward deployed team then costs US$12,000–30,000 a month, depending on its size, for 8 to 16 weeks per workflow.
Where the model comes from
Palantir describes its forward deployed software engineers as engineers who embed directly with customers and build with the people who use the software. Internally it called them Delta, beside the Dev engineers who build the core product.
From 2025 the role spread to AI companies, as businesses tried to turn models into systems that work in daily operations. OpenAI and Anthropic hire forward deployed engineers, and in May 2026 OpenAI launched a separate deployment company, agreeing to acquire a firm of about 150 forward deployed engineers.
Verne adds one discipline to the model: the engagement ends with a handover test your team passes without us.
- A Day in the Life of a Palantir Forward Deployed Software Engineer Palantir Blog, 2 November 2020. Engineers who embed directly with customers and build with the people who use the software.
- Dev versus Delta: Demystifying engineering roles at Palantir Palantir Blog, 8 April 2019. Where the role and the name Delta come from.
- Trading Margin for Moat: Why the Forward Deployed Engineer Is the Hottest Job in Startups Andreessen Horowitz (Joe Schmidt), 4 June 2025. Why AI companies adopted the role.
- Forward Deployed Engineer, Applied AI Anthropic (Greenhouse job board), accessed September 2026. Job posting; describes the role’s intended scope.
- OpenAI launches the OpenAI Deployment Company to help businesses build around intelligence OpenAI, 11 May 2026. Includes the agreement to acquire Tomoro, about 150 forward deployed engineers.
Questions about forward deployed engineering
What is forward deployed engineering?
Forward deployed engineering (FDE) is a way of delivering software in which engineers work inside a client’s operation, learn how the work is really done, and build production systems with the client’s own data and people. The engagement ends when the client’s team can run the system without the engineers.
What does a forward deployed engineer do?
They sit with the people who do the work, measure where the time goes, and write production code against what they find, connecting the systems the business already runs. They also teach the client’s engineers to run what they built, and leave when those engineers pass a handover test.
Where did forward deployed engineering come from?
Palantir used the term for engineers who embed directly with customers and build with the people who use the software. AI companies adopted the role from 2025, as businesses tried to turn models into systems that work in daily operations.
How long does a forward deployed engineering engagement take?
At Verne, a two-week diagnostic comes first. A forward deployed team then works 8 to 16 weeks per workflow, and the engagement ends when the client’s team has run the system for two weeks without help.
How much does forward deployed engineering cost?
Verne’s diagnostic is a fixed fee of US$2,000–5,000. A forward deployed team costs US$12,000–30,000 a month, depending on its size. The build price is fixed in writing at the end of the diagnostic.
When is forward deployed engineering the wrong choice?
When off-the-shelf software already does the job, when the requirements are stable enough to hand to a fixed-scope project, or when nobody on the client’s side will own the system afterwards. In those cases a product, a project or a hire is the better buy.
Who owns the code after a forward deployed engagement?
The client. Verne hands over the code repository and the IP in it at full delivery, along with the runbook and the accounts. On joint work, the client’s engineers get tiered access while the team builds.
Keep reading the guide
Start with one workflow and two weeks
Bring the process that costs you most. The diagnostic measures it, designs the fix and prices the build, and the plan is yours whether or not we build it.