How to hire a forward deployed engineer

One title, several jobs

A forward deployed engineer builds production software inside a customer’s environment and stays until people use it. OpenAI’s San Francisco ad (published September 11, 2026) puts discovery, scoping, system design, build and production rollout in one person, and counts feedback that changes product and model roadmaps as part of success. Palantir’s Forward Deployed AI Engineer ad likens the job to “a hands-on AI startup CTO”.

The title is used loosely. Bloomberry, which analyzed 1,000 postings, says companies use it “to describe three different jobs” (Bloomberry, updated January 25, 2026). By its own classification, which has no stated method:

• Builder, 60% of jobs: “70-90% coding, 30-50% travel”.
• “Sales Engineer+”, 30% of jobs: “30-40% coding, <20% travel”.
• Internal tools builder, 10% of jobs.

Decide which one you are hiring before you write the ad. The builder needs a software engineer’s bar. The second type sits closer to a sales engineer, although Bloomberry found that exactly 0% of the postings it analyzed were quota-carrying.

What to screen for

• Production code and customer delivery, together. OpenAI asks for 5+ years “of engineering or technical deployment experience that includes customer-facing work”, production code across frontend and backend, and experience building or deploying LLM-powered systems (ad published September 11, 2026). Palantir asks for past LLM work, ML basics (“Evaluation, Training, Problem Decomposition”) and strong coding.
• Python. It appears in 66% of the postings Bloomberry analyzed (Bloomberry, updated January 25, 2026).
• Travel. In ads we read on October 11, 2026, OpenAI asks for up to 50% travel, Palantir up to 25% and Databricks 20%. Agree it in the first conversation.
• Platform skills. Databricks’ Sr. Forward Deployed Engineer ad lists “Databricks Certification”. Foundry ads from Booz Allen Hamilton and Accenture Federal Services ask for named Foundry work, such as ontology development, Workshop and data pipelines.
• Clearance for federal work. Booz Allen’s Foundry ad in Arlington, Virginia, requires the ability to obtain a Secret clearance.

How strong teams interview

Few employers publish how they interview for this role. Two public sources help.

Anthropic’s careers page (no date shown) describes its technical interviews in general, not this role: “For technical roles, we use live coding tools like Colab and CodeSignal.” Applicants can look things up, and “We care about what you can do, not where you learned to do it.” Where it makes sense, Anthropic invites applicants to use Claude. The lesson: make the coding round look like the job, and judge the work, not the pedigree.

Palantir’s certification tiers separate knowing a platform from using it. A Palantir Community post of July 27, 2026, describes the Associate certifications as “timed, open-book, multiple-choice exams, valid for two years”, and the Specialist level as “Hands-on, live exams in a real Foundry and AIP environment”. The post does not say whether its author is Palantir staff. For Foundry work, ask which tier an engineer holds, and test the way the Specialist tier does.

From what employers ask for in their ads, a loop that tests both halves of the job has four parts:

1. Hands-on coding on a realistic integration, with lookups allowed.
2. Scoping an underspecified customer problem. OpenAI puts discovery and scoping in the role; Palantir names problem decomposition.
3. A conversation with someone playing the customer, to see how the engineer explains trade-offs and pushes back.
4. For platform roles, a practical check in the platform itself.

Where these engineers come from

Bloomberry looked at 100 LinkedIn profiles of current forward deployed engineers, without saying how they were chosen (Bloomberry, updated January 25, 2026). Their prior roles:

Bloomberry’s reading is that most start as software engineers who develop customer-facing skills, not the other way around. Look first at engineers who have shipped production systems and already spend time with users. Solutions engineers and consultants are worth screening if they still write production code.

Palantir grows its own. It advertises a New Grad forward deployed software engineer role in New York at $135,000 to $145,000 a year, and a “Year at Palantir” internship for the role at $5,900 to $10,500 a month (Palantir, both read October 11, 2026).

Consulting firms are building Foundry benches too. In a press release of July 21, 2025, Deloitte said it has the largest cohort of trained and certified forward deployed engineers outside Palantir and expects to more than triple it. Palantir and Accenture Federal Services said they would train and certify Accenture Federal’s Data & AI team of 1,000 professionals on Foundry and AIP (Intelligence Community News, July 1, 2025). These firms are both a source of trained engineers and a competitor for them.

What employers publish on pay

“Base” means the ad says the range is base salary.

All seven ads were read on October 11, 2026, on each employer’s own careers page. Bloomberry put the median at $173,816 “based on jobs that disclosed salary ranges”, without saying whether that is base (Bloomberry, updated January 25, 2026). Compare offers on the same pay basis. Our forward deployed engineer role page covers the role in more detail.

Contract or permanent

On October 11, 2026, we checked the public job feeds of 62 AI and tech employers. None listed a forward deployed engineer role as contract or temporary. The contract and temporary postings we found were for operations and similar support work. We identified them by each feed’s employment-type field, or by “contract” or an hourly figure in the ad, so a contract role with neither marker would be missed. It is a convenience sample, but it points one way: these employers hire forward deployed engineers as permanent staff.

So plan a permanent hire. If a deployment cannot wait, engineers on contract through their own companies can carry the work while the search runs.

How long it takes

There is no independent data on how long it takes to hire a forward deployed engineer. We found no employer-published or independent time-to-fill figure, only estimates from vendors that sell hiring services or interview preparation, which we do not repeat.

What you control is your own loop: how many rounds, how fast feedback goes back and how quickly you decide. If you run the search with Kernavique, we agree the shortlist timing with you at the brief call, based on the role and the market.

In practice

Write down which version of the role you need, the travel you can offer, the platform and whether the work needs clearance. Build a loop that tests both code and customer work, look first at software engineers who have moved toward customers, and benchmark the offer against the ranges above on the same pay basis. For a permanent search, see Engineering Search.

What we left out

• Hourly figures for contract work. Third-party ads and federal schedules publish them, but they are not employer pay.
• Vendor accounts of specific interview loops and hiring timelines. We could not check them against employers’ own pages.
• Palantir’s own interview method. Its careers pages could not be read as text when we checked.
• What Palantir certification costs and who can sit it. The post does not say.
• Anthropic’s US pay for an individual forward deployed engineer. On October 11, 2026, its job feed listed that role only in Munich, Paris and London.
• How many forward deployed engineers come from Palantir. No source we read measures it.