Forward Deployed Engineer Jobs in San Francisco

The rare FDE seat where the deliverable isn't inference but training itself — embed with AI labs and enterprises to design their post-training runs, RL environments, and eval harnesses on Prime Intellect's Lab stack. Backed by Founders Fund with Andrej Karpathy angel'd in; the $150K floor-to-$300K ceiling spread means they'll level you on research chops, not tenure.

The company selling an autonomous software engineer (Devin) runs a human 'Deployed Engineer' fleet a dozen postings deep — SF/NYC/Austin plus ANZ, Tokyo, Singapore, London, Brazil, South Korea, and a Federal DC seat. They dropped the 'Forward,' but it's the same role, and the tell is unmissable: when the autonomous-coding leader needs this many people to land the product, the last mile is still very much human.

Four FDE postings — including two Regional Director seats in SF and London — from a company being acquired by SpaceX at $60B. Cursor is building a global FDE org mid-acquisition; nobody does that for a side project.

Three ex-OpenAI researchers raised $80M from Benchmark and Sequoia to sell the FDE motion itself — embedding is the product, not the support function. In $1.3B-valuation talks within months of launch.

Mira Murati's lab pays lab-tier comp for the team customers actually meet — fine-tuning as the wedge, FDEs as the delivery mechanism.

Anthropic's FDE org is brand new and scaling fast — early FDEs here will define the playbook.

Ground floor of OpenAI's billion-dollar FDE venture announced in May — the highest-leverage deployment seat in the industry.

The original enterprise-AI pure-play — public since 2020, long before the LLM wave — is hiring FDEs eight postings deep across four cities. The $120K–$158K base sits well under the AI-lab bands: same title, priced by a pre-LLM generation that never treated deployment as a scarce skill. Good volume, honest comp, and a real shot at industrial and public-sector accounts most lab FDEs never touch.

Hebbia sells AI to the buy-side and law firms, and its FDE band tops out at $300K — among the highest published for a non-founding seat. The strict 5-days-in-office rule (NYC or SF, no remote) is the tell: at that comp you're really being hired as the customer's embedded quant.

Distyl is the ex-Palantir crowd running OpenAI's enterprise deployments — 'working AI systems within a quarter' is the pitch. $150–250K with 25–50% travel: this is the Palantir FDE model reborn as an OpenAI channel, staffed hard across healthcare, finance, supply chain and insurance.

The widest band on the board — $113K to $300K in a single req — because Zipline wants a 'startup CTO inside Zipline' who reports to the Head of Operations, not Engineering. FDE is escaping software: here it means wiring GenAI into aviation and logistics for the world's largest drone-delivery network.

Console is pointing state-of-the-art agents at enterprise IT workflows — the least glamorous, highest-ROI corner of automation. An on-site SF FDE at $180K–$280K to prove that the boring internal-ticket stuff is exactly where agents pay for themselves.

Almost every FDE listing right now is an LLM job wearing a deployment hat; this one wants vision-language and visual foundation models pointed at defect inspection and PLC/SCADA integration. The 2x band width ($150K–$300K) says Matroid hasn't decided whether it's hiring a CV researcher or a field engineer — and the JD reads like it genuinely wants both.

Listed in the FDE census, but Nooks has renamed the seat "Agent Engineer" — and then gone further than most FDE roles dare, giving the engineer GM-level responsibility for customer revenue with a 5x meetings-booked target in a one-month trial. Watch this title: when the deployment engineer owns the customer's number, "forward deployed" starts sounding understated.

Lovable scaled to millions of builders with no deployed-engineering org — this is the founding seat, 5+ yrs full-stack (React/TS), embedding with its 'most ambitious customers to translate impossible ideas into real products.' You set the playbooks for FDE at one of the fastest-growing AI app builders.

Exa builds the search engine agents actually run on, and calls its FDE 'the technical tip of the spear' — the person who turns messy enterprise integrations into production wins. A rare deployed-engineering seat on the retrieval layer that sits underneath every agent stack.

Firecrawl turns any website into LLM-ready data and is one of the most-used primitives in the agent stack; $160K–$220K for an FDE shipping customer integrations, with a second broader FDE seat ($150K–$250K) open too — a small YC team standing up deployed engineering as demand outruns headcount.

Reducto's document-parsing platform grew revenue 7x YoY, and this founding FDE owns on-prem/VPC/air-gapped deployments with 'strict latency targets' at $150K–$300K. Doc-AI deals are won or lost in exactly this deployment hell — custom auth, air-gaps, regression evals — which is why the band is so wide.

Retell hit $36M ARR with a team of 20 building voice AI for contact centers and pays 'top 5 percentile' — $215K–$290K for a senior FDE shipping enterprise CRM/ERP integrations. One of the leanest ARR-per-head stories in the census, now hiring deployed engineers to keep pace with enterprise demand.

Edison Scientific's Kosmos 'AI scientist' runs hundreds of research tasks in parallel, compressing months of R&D into one run; its forward-deployed AI engineer ($200K–$350K) embeds on-site with a top science org — explicitly 'not remote support or implementation consulting.' After Normal Computing's silicon play, the rarest vertical in the census: FDE for frontier scientific discovery.

Figma is standing up an FDE motion around Make Local, its new design-to-code developer workflow — embedding engineers with strategic customers to unblock adoption and feed the product. When a design giant with millions of users spins up forward-deployed engineering, the title has fully crossed into mainstream software. $152.8K–$296K is a wide leveling band for a function it's building from scratch.

P-1 AI is building 'Archie,' an AI engineer agent for the physical world (mechanical/electrical/aerospace design), and its FDE is the technical lead who wires Archie into real engineering workflows and fine-tunes it on customer data. It explicitly wants a physical-engineering background — one of the only FDE seats that asks for aerospace/mechanical chops, not just backend. Rarest vertical on the board after silicon and drug discovery.

A robotics-automation startup building its FDE bench three postings deep — Robotics, Vision & ML, and an Edge-Frameworks/FDE-tooling seat — at $160K–$200K + equity to deploy FactoryOS on real factory floors. This is the corner of the market where 'forward deployed' means a hard hat, not a Zoom call.

A $145M-funded edge-sync database trusted by Delta, Chick-fil-A, and the US military, now hiring FDEs five postings deep including a DC/defense seat. The work is genuinely hard — peer-to-peer CRDT sync across offline and intermittently-connected devices — which is exactly why it's FDE-shaped rather than a solutions-engineering rebrand.

A bio-AI FDE: you're the technical face to biopharma partners, wiring Chai's molecular-interaction models into wet-lab and in-silico workflows. Proof the FDE motion has reached drug discovery — a vertical almost no other FDE has touched.

A 'Founding' Forward Deployed ML Engineer at a young lab hiring across SF, Singapore, and India at once — founding scope on the customer-facing edge, where you 'operate at the bleeding edge of our technology and our customers' most consequential problems simultaneously.'

Founded by the Stanford AI Lab PhDs who invented State Space Models — now embedding FDEs to push agentic voice into on-prem and VPC enterprise stacks. $180K–$250K + equity for 4 years' backend is mid-band, but the SSM pedigree and Databricks/Index/Lightspeed cap table are the real comp.

$300M+ ARR, 5x growth, a $4.65B Series C — and the FDE team ships open source (they contribute to SGLang) while tuning inference for Suno, Lovable, and Cognition. Only 2 years' ML asked; the job is GPU-level performance work at frontier labs, not slideware.

The rare FDE that's hands-on-hardware: you're the first responder for a live robot fleet at one of the hottest robotics foundation-model labs, not embedding at a customer. Asks just 2 years and a B.S. — because almost nobody has actually debugged multi-robot fleets in production.

An $11B, $781M-raised voice-AI leader running a fully remote-global FDE seat — rare for a function usually chained to on-site enterprise work. The paper bar is unusually low ('customers in student clubs count'); the real bar is high: you own end-to-end builds for their most strategic accounts.

The only FDE seat where you live inside SystemVerilog and UVM testbenches — Normal calls itself a 'frontier AI lab for silicon,' embedding engineers with the world's largest chipmakers to cut EDA cost 1,000x. $180K–$325K to own a vertical almost no other FDE can touch.

A generative-video FDE seat that turns Luma's models into production systems on a customer's data — and only asks for 2+ years, with an explicit 'we're not looking for unicorns' pitch that quietly undercuts the usual FDE bar. $170K–$290K, embed-with-the-customer model.

$200K–$400K is the widest band in this week's mid-cap batch — and it's Meticulous's first US FDE seat, establishing the function from scratch. Their AI writes test suites automatically; the customer list (Notion, ElevenLabs, Dropbox, Discord, CoreWeave) is the pitch.

The company running Anthropic's $50B compute buildout pays $150K–$275K for an FDE who deploys inward, not outward — no customers, just their own gigawatt-scale stack. The clearest sign yet the title now means 'engineer with maximum autonomy.'

America's open-frontier-lab bet — $2B raised, reportedly seeking $25B — is hiring FDEs five postings deep, including a Seoul seat and an 'LLM post-training' FDE variant no other lab offers.

Four GTM-FDE seats across SF, Tokyo, Munich, and Paris — including one of the only FDE roles in Japan anywhere in the 1,206-posting Plank census (CC BY 4.0).

The FLUX team's forward-deployed seat — proof the FDE motion has spread from LLM labs to image gen. Requires actual diffusion-model internals knowledge, which shrinks the candidate pool to near zero.

The most precise comp band in the 1,206-posting Plank census — a $20K spread where everyone else posts $200K ranges. Anyscale knows exactly what this seat is worth; Ray sits under half the industry's training stacks.

Four parallel 'Forward Deployed Agent Builder' postings across SF, NYC, Vancouver, and São Paulo — Brex is standing up a global FDE function, and the non-US seats are among the only LatAm/Canada FDE roles anywhere.

Baseten is hiring the FDE manager and the ICs simultaneously — join the org on day zero, or run it.

A seed-stage company posting a $500K ceiling — the second-widest band in the 1,206-posting Plank census after Tenex. Either desperation or conviction; both are worth an interview.

Bret Taylor's agent company invented the modern agent-deployment motion — the infra variant of the role is rarer and pays accordingly.

Rippling's FDE function is brand new — join early enough to set the playbook rather than follow it.

Eval and data-quality work is the most durable skill in the FDE toolkit — Labelbox FDEs do it all day.

Contact-center AI is one of the few categories with proven ROI — Cresta FDEs ship deployments that visibly move customer P&L.

Scale's Donovan and enterprise GenAI work spans defense and Fortune 500 — unusual breadth, and a path to cleared work that pays a premium.

Fin is one of the highest-volume production AI agents in the world — FDEs here optimize systems handling millions of conversations.

Stripe engineering brand plus customer-facing autonomy — FDEs here work with the largest platforms integrating payments and AI.