Vibe Engines is free, interactive AI engineering & system design — runs in your browser. Pick a learning path

Forward Deployed Engineer / The FDE Canon

FDE Track~12 min readUpdated Aug 2026

Reading shelf

Ten sources,
each wrong about something

Most writing about forward deployed engineering is recruiting copy or speculation. A small number of pieces are genuinely load-bearing: first-hand accounts by people who did the job, primary documents worth reading as documents, and market data. Here is the shelf, in reading order, with what to take from each — and what each one is silent about, because that matters as much.

Start here: what the job feels like

Two first-hand accounts. Read these before any job description, or you will read the job descriptions wrong.

Nabeel Qureshi — “Reflections on Palantir”¹⁴

Take: the texture of forward-deployed work at the company that invented the title — the Airbus engagement, the culture, what it feels like to be dropped into an organisation and expected to produce. It is the best answer available to “what is this actually like?” and it is unusually honest about the parts that were bad.

Silent about: the modern AI-native version of the role, and compensation. It describes one company, one era, one person’s experience — which is its strength as a document and its limit as a guide.

PostHog — “WTF is a forward deployed engineer?”¹⁵

Take: the same role at a product company rather than a platform company, written first-party and plainly. It is the clearest available account of the travel reality, which most sources skip and which is the single most common reason people leave the role.

Silent about: enterprise-scale deployment concerns — security review, air-gapped environments, the commercial layer. A smaller-company view, and the more relatable one.

Baseten — “What I learned as an FDE at an AI startup”³¹

Take: the AI-native version, at the stage where nobody has patterns yet. Useful precisely because it is not from a mature FDE organisation — it describes the improvisation that a polished job description hides.

Silent about: what scales. One engineer’s tour at one startup; read it for the shape of the days, not the method.

→ Why first-hand accounts come first

Job descriptions describe the role a company wants to hire. First-hand accounts describe the role as lived, including the parts nobody would put in a posting: the travel, the loneliness of being the only engineer who understands the system, the weeks where you are a project manager with a laptop. Reading the postings first calibrates you to the marketing.

Primary documents: read the job descriptions as documents

Not to apply — to see what companies are committing to in writing.

Source	What to take	What it is silent about
OpenAI — FDE job description⁹	The phrase “full arc of a deployment”. It is the clearest single statement of the builder variant, and it is what to hold a company to in an interview.	How many customers at once, and whether any of the comp is variable. Both are askable.
Anthropic — FDE, Applied AI¹⁰	A published 40/30/30 split and MCP servers, sub-agents and agent skills named as deliverables. The 30% that feeds back into the product is the part that distinguishes this from consulting.	What happens when the 40% expands to fill the quarter, which is the normal failure mode of that split.
Anthropic — FDE, Federal Civilian²⁹	Air-gapped and impact-level deployment stated as requirements. Read it if you want to understand the regulated end of this work — it is a different job.	The personal cost of clearance work: travel constraints, portfolio invisibility, on-site requirements.
Palantir — “Dev versus Delta”¹³	The two-role split — FDSE and Deployment Strategist — that most companies copied the title of without the structure. It explains why one word now means two jobs.	That almost nobody else adopted the pairing, so a single FDE elsewhere absorbs both accountabilities.

→ The trick with job descriptions

Read them for what a company is committing to rather than what it is asking for. “Owns the full arc of a deployment” is a promise about scope; a published time split is a promise about how your quarter is allocated. Both are legitimate things to ask about in an interview, and asking signals that you read carefully — which is the trait the job is about.

The commercial argument, and the market data

a16z — services-led growth⁸

Take: the reason the role exists, stated as a trade — margin for defensibility. If you understand this essay you can predict most decisions your employer will make about your team.

Silent about: the failure mode it walks into. It argues for the trade and does not dwell on organisations that pay the margin and never receive the moat. Read alongside our breakdown.

MIT NANDA — the 95% figure¹⁷

Take: the demand-side explanation. Capability is available and outcomes are not, which is a deployment gap rather than a model gap.

Silent about: its own methodology — the figure circulates through press coverage rather than a broadly available paper, so quote it as approximate. Our breakdown handles the caveats.

Bloomberry — 1,000 FDE jobs analysed¹

Take: that one title covers three different jobs, and roughly a third of postings are pre-sales-shaped. This single fact will save you from accepting the wrong role.

Silent about: filled roles. Posting counts measure demand signals, and title inflation is real in a hot market. Breakdown here.

First Round Review — hiring an FDE¹¹

Take: the hiring side, which tells you what is actually being assessed. It puts the engineering bar at staff level plus a temperament most strong engineers have not been asked to develop — useful because it names the second half explicitly.

Silent about: what the role does to a career over five years, which is the question candidates should be asking.

The Pragmatic Engineer — forward deployed engineers¹²

Take: the industry-wide framing, written for engineers rather than for investors or recruiters. The best single orientation piece if you read only one.

Silent about: the practitioner detail — evals, identity, the security review — which is where the work actually lives.

Where the role is going

Latent Space — forward deployed engineers at the AI Engineer World’s Fair³⁰

Take: the 2026 shift some teams describe — from writing feature code to writing the systems that write feature code, with engineers supervising agent harnesses. If that direction holds, it changes what the deliverable is, and the agent harness design is our attempt at what it implies.

Silent about: whether it generalises. One company’s direction of travel, described at a conference, is a hypothesis rather than a trend.

Palantir blog — a day in the life³³

Take: the concrete shape of a week at the origin company, which is a useful anchor against which to read every newer description.

Silent about: everything after 2020-era Palantir. Read it as a historical baseline, not a current spec.

→ What is missing from this shelf, honestly

There is very little good writing on the parts of the job that consume the most time: enterprise identity, the security review, the commercial layer, and evaluation for client work. That is not an accident — those topics are unglamorous, company-specific and rarely written up by the people doing them. It is also why most of this site exists. If you know of a genuinely good first-hand piece on any of them, it belongs here.

After the reading

The reading is the cheap part

Every source above describes the same underlying skill: sitting with ambiguity in front of someone who is paying, and producing a next step rather than a caveat. That is practisable.

The full roadmap Decomposition simulator Client roleplay How to become one

The picture

The picture: the shelf in reading order

Read left to right. Each shelf calibrates the next: first-hand accounts tell you what the job descriptions leave out, the job descriptions tell you what is promised in writing, and the market pieces explain why anyone pays for it. The fourth shelf is a direction, not a fact. The dashed shelf is where most of the working time goes and where almost nothing good is written.

By the numbers

Worked example: reading one posting as a document

Apply the page’s method to a posting that uses the phrases quoted above. For each line, ask what it commits to and what it leaves out. The quarter is 13 weeks; the overrun in step 3 is an illustrative scenario.

Step	The posting says	What it commits to	What to ask
1	“Owns the full arc of a deployment”	Scope from first meeting to running in production	Who owns it after go-live, and for how long?
2	A 40/30/30 time split	Over 13 weeks: 5.2, 3.9 and 3.9 weeks	Which share gives way when the 40% overruns?
3	(Scenario) the 40% grows to 60%	7.8 weeks, 2.6 more than promised	If it all comes out of the product-feedback 30%, that share falls from 3.9 to 1.3 weeks: two-thirds gone
4	Nothing on customer count	Silent	How many customers at once?
5	Nothing on pay structure	Silent	Is any of the comp variable?
6	Air-gapped or impact-level deployment	A regulated job	What do travel, on-site work and clearance cost me?
7	Quota, demos, “partner with sales”	Pre-sales-shaped, like roughly a third of postings	Is this the builder variant at all?

Step 3 is why the page calls the 30% that feeds back into the product the part that separates the role from consulting. It is also the share that disappears first under customer pressure. Asking steps 2 to 5 in an interview costs nothing. The answers tell you which of the three jobs behind the title you are actually being offered.

Trade-offs

Trade-offs: what each kind of source costs and buys

Here Time is the effort to read it, Space is what it covers, and Why is the bias to correct for.

Case	Time	Space	Why
First-hand accounts	One long read each	Texture: travel, isolation, the bad weeks	One person, one company, one era.
Job descriptions	Minutes per posting	What is committed to in writing	Describes the hire a company wants, not the job as lived.
Investor essays	One essay	The economics: margin traded for defensibility	Argues for the trade; skips the firms that pay and get no moat.
Market data	Fast to quote, slow to check	Demand signals at scale	Postings are not filled roles; check how a headline figure was produced.
Hiring guides	One read	What is actually assessed	Written for the employer, silent on your five-year arc.
Conference talks	One talk	Where some teams are heading	One company’s direction is a hypothesis until others follow.

Watch for

Common mistakes

Reading postings first
Job descriptions calibrate you to recruiting language. Read two first-hand accounts first, then the postings, and the gaps between them become your interview questions.
Quoting the 95% as a precise fact
The figure reaches most readers through press coverage, not the method. Say “roughly 95% of pilots, in one widely reported study” and use it as a direction, not a statistic to build a case on.
Counting postings as jobs
Posting counts measure demand signals. In a hot market titles inflate, and some postings are never filled. Growth in postings is not growth in people doing the work.
Treating one title as one job
The same word covers builder, pre-sales and internal roles. Accepting an offer on the title alone is how people end up in a sales role they did not want; step 7 above is the check.
Reading a baseline as a current spec
The Palantir pieces describe one company in an earlier era. Use them as an anchor to compare newer descriptions against, not as a description of the job at an AI company today.

In the wild

Where this shows up: interviews and the first month

In the interview
The steps 2 to 5 questions show the careful reading the role is hired for. “How many customers at once?” and “what gives when the 40% overruns?” are fair, specific and rarely asked.
Choosing between offers
Put each posting through the same seven steps. Two offers with the same title can turn out to be a builder role and a pre-sales role once you look at quota language and the time split.
Explaining the role to a client
The a16z trade is the short answer to “why is an engineer from the vendor sitting with us?”: the vendor spends margin on deployment depth because depth is what keeps the account.
Filling the missing shelf
Identity, the security review, the commercial layer and client evals are learned on the job. Writing up what you learn in your first month, with client details removed, is the scarce document every newcomer wishes existed.

Self-check

Check yourself

A friend has two hours before an FDE interview. What should they read, in what order?

One or two first-hand accounts (Qureshi, PostHog) for the job as lived, then that company’s own posting read as a document. The gaps between the two become their questions. Market data can wait: it explains the industry, not this role.

A posting promises 40/30/30, and the hiring manager says the 40% “sometimes runs over”. Over a 13-week quarter, what is at stake if it reaches 60%?

7.8 weeks instead of 5.2, so 2.6 extra weeks must come from somewhere. If they come from the product-feedback 30%, it falls from 3.9 to 1.3 weeks, and the role drifts towards consulting. Ask which share gives.

A slide says “FDE postings are booming, so FDE employment is booming.” What is wrong?

Postings measure demand signals, not filled roles, and title inflation is real in a hot market. Roughly a third of postings are also pre-sales-shaped, so even the postings are not all the same job.

A conference talk says engineers will soon supervise agent harnesses instead of writing features. How should you weigh it?

As a hypothesis. It is one company’s direction, described at a conference. Take the implication seriously (the deliverable may change) but wait for other teams to report the same before planning a career on it.

Frequently asked

Quick answers

What should I read first to understand forward deployed engineering?

First-hand accounts before job descriptions. Nabeel Qureshi’s "Reflections on Palantir" for the texture of the work at the company that invented the title, PostHog’s first-party account for the travel reality that most sources skip, and Baseten’s account for the AI-native version at a stage where nobody has patterns yet. Reading postings first calibrates you to recruiting language rather than to the job.

Why read FDE job descriptions as documents?

Because they show what a company is committing to in writing, not just what it wants. "Owns the full arc of a deployment" is a promise about scope; a published time split such as Anthropic’s 40/30/30 is a promise about how your quarter is allocated. Both are legitimate to ask about in an interview, and asking demonstrates the careful reading the job itself requires.

What explains why demand for the role grew?

Two sources together. a16z’s services-led-growth argument gives the supply-side logic — deliberately trade gross margin for deployment depth because depth is defensible. The widely reported finding that roughly 95% of enterprise AI pilots showed no measurable P&L impact gives the demand-side logic: capability is available and outcomes are not, which is a deployment gap rather than a model gap.

What is missing from the available writing about this role?

Almost everything about the parts that consume the most time: enterprise identity and SSO, the vendor security review, the commercial layer of SOWs and pricing, and evaluation for client work. Those topics are unglamorous, company-specific and rarely written up by the people doing them, which is why most practitioner material has to be assembled rather than found.

Where is the role heading?

One direction described at the 2026 AI Engineer World’s Fair is a shift from writing feature code to writing the systems that write feature code — engineers supervising agent harnesses rather than authoring every change. Treat it as a hypothesis rather than an established trend: it is one company’s direction of travel, described at a conference, and it has not yet been shown to generalise.

Receipts

Sources

Every number on this page traces to one of these. Where a figure is self-reported or crowd-sourced rather than first-party, it is labelled inline.

Cited on this page

Bloomberry — “I analyzed 1,000 forward deployed engineer jobs” (Nov 2025) — posting growth, builder/pre-sales/internal split. bloomberry.com/blog/i-analyzed-1000-forward-deployed-engineer-jobs-what-i-learned/
a16z — Services-Led Growth (Joe Schmidt, 4 Jun 2025) — the margin-for-moat thesis behind FDE hiring. a16z.com/services-led-growth/
OpenAI — Forward Deployed Engineer job description — “full arc of a deployment”. openai.com/careers/forward-deployed-engineer-(fde)-nyc-new-york-city/
Anthropic — Forward Deployed Engineer, Applied AI (JD) — 40/30/30 split; MCP servers, sub-agents and agent skills as deliverables. jobs.menlovc.com/companies/anthropic/jobs/69674588-forward-deployed-engineer-applied-ai
First Round Review — “So you want to hire a forward deployed engineer” — the staff-engineer bar. review.firstround.com/so-you-want-to-hire-a-forward-deployed-engineer/
The Pragmatic Engineer — Forward deployed engineers — newsletter.pragmaticengineer.com/p/forward-deployed-engineers
Palantir blog — “Dev versus Delta” — FDSE and Deployment Strategist as distinct forward-deployed roles. blog.palantir.com/
Nabeel S. Qureshi — “Reflections on Palantir” — first-hand account of forward-deployed work, including the Airbus engagement. nabeelqu.co/reflections-on-palantir
PostHog — “WTF is a forward deployed engineer?” — first-party account of the role at a product company, including travel reality. posthog.com/blog/forward-deployed-engineer
MIT NANDA “State of AI in Business” — 95% of pilots with no measurable P&L impact — report dated July 2025, available as a PDF; the figure here is cited via later press coverage (TechCrunch, 30 Jul 2026), which is what this link points to. techcrunch.com/2026/07/30/forward-deployed-engineers-are-the-ai-industrys-latest-talent-obsession/
Anthropic — Forward Deployed Engineer, Federal Civilian (JD) — air-gapped and IL-level deployment requirements. jobs.menlovc.com/companies/anthropic/jobs/76278352-forward-deployed-engineer-federal-civilian
Latent Space — Forward deployed engineers (AI Engineer World’s Fair) — Ramp’s shift toward engineers supervising agent harnesses. www.latent.space/p/forward-deployed-engineers-aiewf
Baseten — “What I learned as a forward deployed engineer at an AI startup” — www.baseten.co/blog/what-i-learned-as-a-forward-deployed-engineer-working-at-an-ai-startup/
Palantir blog — a day in the life of an FDSE / Deployment Strategist — blog.palantir.com/

Finished this one? 0 / 209 Handbooks done

NextFDE Interview Prep: The 9 Rounds Next in “Become a Forward Deployed Engineer” · step 52 of 63

Explore the topic

See this alongside everything else on the same subject — handbooks, system designs, challenges and tools, in one place.

Interview Prep Forward Deployed Engineering

More Handbooks

Explore more

Labs →

Stakeholder Politics Lab

Paper Breakdowns →

The 95% Number, Examined

Challenges →

Deep Clone, Constraint by Constraint