What a Forward Deployed Engineer really is

What an FDE is

A Forward Deployed Engineer is an engineer embedded with a customer who turns a capable AI product into a working solution inside their environment. The model was pioneered at Palantir — working with users who could not articulate what they needed, engineers showed rough demos and iterated on the feedback — and has since been adopted, almost word for word, by the frontier AI labs. It is a senior engineering role with two axes most engineering roles never touch: customer engagement and ownership of the outcome. Three traits define it.

Trait percentages are from our own scan of 1,190 US FDE job postings, August 2026.

“Isn’t this just a glorified Solution Architect?”

It is the first thing most engineers say when they hear the title, and the skepticism has an honest basis. The title began as a rename. In 2011 Palantir rebranded its solutions engineers as Forward Deployed Engineers — the venture firm a16z calls it the canonical case of “title arbitrage.” So the doubters are right about where the name came from. What they miss is what changed underneath it.

a16z proposes a useful test for whether a new title is real: “does this describe work that would have been unrecognisable five years ago?” For the FDE it does. The thing being deployed is no longer deterministic software whose requirements can be specified up front — it is an AI system whose capabilities only become clear once you build against real customer data. That breaks the classic sequence of architect-designs → someone-else-builds, and collapses it into one person who does both, on site. Here is how that lands against the three roles the FDE is most often confused with.

And it is not consulting

The consulting comparison is the one buyers themselves push back on hardest. The distinction is structural, not tonal: a consultancy sells hours against a scope, while an FDE ships product-backed software and is measured on whether it runs. In McGrew’s framing, FDEs lay “gravel roads” — one-off custom builds for a single customer — which the central product team then paves into “superhighways,” generalised platform features. That feedback loop into the product is what stops the motion collapsing into consulting, and no consultancy has it.

“We didn’t need more consultants, we needed engineers who could build what didn’t exist yet.”

— Bala Vadhiyar, CTO, JPMorgan Chase · on Databricks’ forward-deployed engagement (Jun 2026)

Sources: a16z on forward-deployed job titles and the 2011 Palantir rebrand (Jan 2026); Bob McGrew’s FDE playbook (Y Combinator); Databricks’ Forward Deployed Engineering announcement (Jun 2026).

Want the full head-to-head — skill shape, output, ownership? See FDE vs Solution Architect →

Why the role exists now

The FDE appeared because the value gap in enterprise AI is enormous — and it sits in deployment, not in the model. Capable models are widely available; getting one to run dependably inside a real business is the hard, unglamorous, well-paid part.

As Forbes put it, “the most expensive job in enterprise AI is no longer the researcher building frontier models but the engineer flying to a customer site to make those models work” (May 2026). The labs agree with their capital. In May 2026 OpenAI stood up a dedicated “Deployment Company” — a $4B+ venture whose entire job is to embed Forward Deployed Engineers into enterprises and turn pilots into running systems.

What an FDE actually does — the engagement arc

The FDE is a delivery-owning role: once a deal is real, it runs the whole technical arc, from its own discovery through to a system the customer’s team can run without it. This is the shape of the job.

1. 01
   Technical discovery & scoping

   Shadow the users, map the real workflow, and frame the smallest valuable first deployment. This is the FDE’s own scoping — technical (“how do we build this in your stack?”), not the commercial qualification a sales team runs.

2. 02
   Architecture

   Design a solution that fits the customer’s data, systems, and security constraints — not a greenfield ideal.

3. 03
   Build & integrate

   Write and ship the production code, wiring the AI into the customer’s existing systems and data.

4. 04
   Deploy

   Put it live in the customer’s environment — often regulated, on their own infrastructure.

5. 05
   Handover

   Leave clear docs, runbooks, and a customer team that can run and extend it without you.

6. 06
   Expand

   Turn the first win into the next, and feed what you learned back to the product team.

The FDE rarely runs this arc alone. In a typical enterprise engagement the pod is a deployment strategist (drives the meetings, the architecture, and engagement management), one or more FDEs (technical evaluation, implementation, demos on shorter milestones), and an integration FDE pulled in when the customer’s data sits across multiple sources (MCP connections). The arc also has an exit built in: the FDE lands the first agent in production to prove adoption, then rolls off to a larger professional-services team — the FDE owns the proof of value, not the multi-year rollout.

Expect a bigger communication load than the title suggests. In enterprise FDE practices, roughly 30% of the role is executive communication — selling the vision and convincing customer stakeholders to adopt agents. That is not a sales job in disguise: sales hands post-deal work to professional services, while the FDE stays hands-on through the first deployment.

Every stage above maps to a skill set — that is the next guide. Walk the full FDE skills roadmap →

One title, many shapes

The most common misread is treating “FDE” as a single job at a single level. It is not. The title spans a full seniority ladder, several domain specialisations, and a set of adjacent titles.

It is a ladder, not an IC job

Forward deployment is being built as an organisation, not a handful of hires. OpenAI, Google Cloud, Microsoft, Amazon, Anthropic and Databricks have each stood up a dedicated forward-deployed org — Microsoft’s is the largest single bet at $2.5B and 6,000 embedded experts, while Databricks reached the same place from the other direction, consolidating its existing Professional Services delivery arm under the FDE banner. Colin Jarvis, who leads OpenAI’s forward-deployed org, grew his team from 2 to roughly 52 engineers in a single year and calls it “the hardest job to hire for.”

Only two of them have published enough structure to show a complete ladder — so those are the two below. Read them as representative of the pattern, not as the only two that have one.

OpenAI — the forward-deployed org

Google Cloud — the FDE org

It specialises by domain

Beyond seniority, the role bends toward whatever the employer deploys. These are all real posted variants — the same embedded motion with a different centre of gravity.

And it varies by who employs you

The same title means three quite different jobs depending on whose payroll you are on — worth knowing before you target a search.

Two adjacent titles worth knowing

Some orgs split the role by whether it leans build or sell. Meta runs both a Forward Deployed Engineer (build) and a Forward Deployed Solutions Architect (pre-sale, advisory); NVIDIA’s equivalent is a Forward Deployed Architect. Most FDE-titled roles are build-leaning — but read the posting, not the noun.

Also watch employer-specific names for the same seat: Salesforce titles its FDEs “AI Deployment Strategist (FDE)”, and at Google the Customer Engineer / Outcome CE role is the closest sales-engineer equivalent, with Forward Deployed Applied AI focused on Cloud-stack adoption — voice agents and call-center automation. Search those titles too.

Platform experience is rarely the gate. Enterprise FDE practices hire from Azure, LangChain, and LangGraph backgrounds with no prior experience of the employer’s own platform — the foundational agentic/LLM skill set transfers, and any of AWS, GCP, or Azure is acceptable.

One more axis: pay is bimodal. Enterprise FDEs sit near the senior SWE/ML line — around $211K total comp — while frontier-lab FDEs run far above it, $560K+. Same role, two very different tiers, depending on who you deploy for.

Want pay broken down level by level, against the ML and AI Engineer? See the three-role career map →

The same role, across every major lab

This is not an OpenAI-and-Google story. Every frontier lab — and the enterprise platforms and consultancies around them — now runs a forward-deployed ladder, and the biggest names in AI have put ~$9.75B directly behind it (OpenAI $4B+ · Microsoft $2.5B · Anthropic $1.5B · Amazon $1B · Google $750M). The tell is not any one hire; it is that the same ladder keeps appearing, backed by real capital.

Comp bands and levels are drawn from our own lab deep-dives — OpenAI, Google and Anthropic link to the full breakdown. For who is hiring the title at scale, the geography, and the seniority mix across 1,190 postings, see the job-market guide.