The number
6,000. That's how many industry, engineering, and AI experts Microsoft is staffing into "Microsoft Frontier Company," the $2.5B operating business it launched on July 2 to embed its own people inside customers and build their AI systems for them. It landed two days after AWS committed $1B to the same model, and months after OpenAI and Anthropic stood up their own deployment arms. Add it up and two hyperscalers alone put roughly $3.5B behind forward-deployed engineering in a single summer week — not as a pitch, but as a P&L line with a named CEO and a headcount target. The FDE isn't the "hottest job in AI" because a blog said so. It's because the largest software companies on earth just decided the deployment gap is worth billions to close in person.
data: Microsoft Frontier Company launch — CNBC, TechCrunch, GeekWire (Jul 2, 2026); AWS $1B FDE commitment — Channel Dive (Jun 30, 2026).
What changed this week
After nine-plus weeks frozen, the Plank FDE census finally re-ran its crawl — and the movement is real signal, not noise.
The headline count dropped 1,206 → 982 (−19%), but that's a cleanup, not a contraction. The vanished names are overwhelmingly renames (
AI Fund→AIFund,Catena Labs→Catena) and staffing/consultancy evergreen noise (Accenture, Capgemini, back-office roles) aging out — exactly the aggregator clutter that never belonged in a pure-FDE count. Read the −224 as a de-staling pass that makes the remaining number more honest.Anthropic went 1 → 7 in a single crawl — the biggest real mover in range. The census now lists seven distinct Anthropic FDE reqs: FDE roles in NYC, Paris, and Munich; Head of FDE in SF; FDE Managers in London and NYC; and a Pre-Sales Program Lead for FDE. When a frontier lab adds a Head of, two Managers, and its own pre-sales program lead at once, it has stopped hiring FDEs and started building the org that hires them. Mid-market echoed it: Beacon 1→4, Ashby 1→3, Baseten 2→3, Cape 2→3.
Not everything grew. Braze collapsed 8 → 1 and Candid Health 3 → 1 — the sharpest pullbacks in range. Worth watching whether that's a reclassification or a genuine retreat from the embedded model.
data: Plank FDE census (CC BY 4.0). Note: this run's snapshot transferred only the alphabetical head (A→Clera range) due to an upstream fetch-size limit, so full-tail moves aren't visible; count changes above are for exactly-matched companies in that range.
New roles
DeployCo (OpenAI Deployment Company) — Founding Forward Deployed Engineer San Francisco / New York (hybrid, up to 50% travel) · $170K–$400K + equity The $4B OpenAI deployment venture is now hiring its founding FDEs — a ground-floor seat at the entity built to be the tip of the spear for enterprise GPT rollouts. The band spans nearly 2.4x from floor to ceiling, which tells you they'll calibrate hard on the deployment track record you walk in with. View role →
Avoca — Forward Deployed Product Engineer New York, NY (on-site HQ, travel to customer sites) · $160K–$300K + equity + bonus Note the title: Forward Deployed Product Engineer. Series B, $1B valuation, and a JD built to wire what FDEs learn in the field straight back into the product. If you want the FDE lifestyle without giving up product ownership, this is the cleanest version of that hybrid we've listed. View role →
AI Aspire (Andrew Ng / AI Fund) — Forward Deployed Engineer Mountain View, CA (on-site, 1–3 week client engagements) · $180K–$200K The DeepLearning.AI founder is building an FDE bench, and the JD explicitly wants fluency in Claude Code and Codex — a tell that agentic coding tools are now a screened FDE skill, not a nice-to-have. Short, sharp client engagements; a rare tight, disclosed band with a marquee name attached. View role →
Cinder — Forward Deployed Engineer New York, NY (hybrid; will relocate the right person) · $165K–$235K + equity Accel/YC-backed, built for Trust & Safety — an FDE seat inside a mission vertical rather than a sales-adjacent one. Proof the role keeps migrating into domains where the deployment problem is content and risk, not just data plumbing. View role →
Clearly AI — Forward Deployed Security Engineer Remote (US; Seattle/SF preferred) · $125K–$175K + 0.1–0.5% equity Watch the title mutate: "FDE" is becoming a prefix. This YC AI-security startup wants a Forward Deployed Security Engineer, the same way we've now seen Forward Deployed Product and Forward Deployed Software Engineer. The specialization is happening at the title level, not just the JD. View role →
Legion Intelligence — Senior Forward Deployed Engineer (D.C. Area) Washington, D.C. area (on-site/embedded, 80–100%; classified environments) · $160K–$200K Another cleared FDE seat that isn't Palantir — embedded 80–100% inside classified environments. The federal FDE lane we've tracked all summer keeps widening past the incumbents, and the clearance is doing the same work on comp as always: cash-heavy, geo-locked, near-total job security. View role →
Picarro — Forward Deployment Engineer, Oil, Gas & Chemicals Remote (US; CA or TX preferred, up to 70% travel) · comp not disclosed The FDE model has left software entirely and landed in heavy industry. Picarro embeds you at oil, gas, and chemical customer sites — up to 70% travel — to deploy emissions and process-monitoring systems. When a sensors-and-analytics company adopts the FDE title, the category has officially escaped its tech origins. View role →
One insight: the one interview round that actually decides FDE offers
If you're a strong backend engineer getting FDE rejections, it's probably not the coding round that's sinking you. The FDE loop looks like a normal SWE loop — recruiter screen, one or two coding rounds, a technical deep-dive — but the round that actually decides the offer is the ambiguous case study, and it's the one candidates prepare for least.
Here's the shape of it: the interviewer hands you a deliberately under-specified brief — a confused stakeholder, a vague business complaint, no ticket — and watches how you turn it into a buildable plan. They're grading whether you ask the questions that surface the real problem behind the stated one, whether you can defend a scope cut, and whether you can explain a tradeoff to a non-technical budget-holder without dumbing it down. OpenAI's loop reportedly weights case studies, customer empathy, and business judgment at roughly half the evaluation. At a product company the ticket is the input; in an FDE seat, producing the ticket is the job.
A reader story makes the point better than any framework. Daniel Monzon, an FDE, described a client phone port that ran about eleven weeks end to end — five of them just from submission to live. The actual technical cutover took roughly ten minutes. The other eleven weeks went to a production-and-ops slog, including a fraud-ops restriction — tripped by a $40 account upgrade — that quietly routed verification emails to an inbox nobody was watching and stalled the whole port. The cutover was fast because he'd rehearsed it on a $1 throwaway number first. And the honest part he wanted kept in: that rehearsal wasn't clever foresight. He was blocked, impatient, and bought the throwaway just to prove the voice-agent call path worked while the real port sat stuck — the rehearsal was a side effect. His takeaway, "de-risk the part you don't control, early and cheaply," is the thing he says he'd now do on purpose.
That's the muscle the coding round can't test and the case study is hunting for: not can you write the cutover, but can you survive the eleven weeks around it. (Thanks, Daniel.)
data: FDE interview structure and evaluation weighting per Exponent (Aced) FDE interview guides and Perspective AI, "Forward Deployed Engineer Interview Questions: 2026 Prep Guide." Reader story shared by Daniel via LinkedIn, used with permission and attributed by request.
One ask
Forward this to one engineer you know who's eyeing an FDE move — the interview breakdown above is exactly what most of them are prepping for wrong. Every forward is how this list finds the next reader who needs it.