Every stalled AI pilot is missing the same role. Build your Forward Deployed Engineers here.
Forward Deployed Engineers sit inside the workflow, surface the problem the customer actually has, build the fix, and graft it onto the systems already running. Nuvepro is where you build that team. The AI portion and the human portion, both.
641 FDE postings · 214 distinct titles · from 264,613 real job postings · see the live data →
What a Forward Deployed Engineer actually is
Palantir invented the title: engineers stationed inside the customer's operation, close enough to the work to see what a demo never shows. The rest of the market has now adopted it. Sits inside the customer's workflow. Palantir-originated; Indeed reported Forward Deployed Engineer postings grew more than 700% year over year. Solutions-architect-flavored: bridges model lab and customer, deliverable is a tailored integration plus handoff.
What separates an FDE from a solutions engineer is ownership of both halves. A solutions engineer scopes and hands off. An FDE discovers the real problem, builds the fix, and stays until it runs inside the customer's systems. The deliverable is not a recommendation. It is a working change in how the work runs.
Andrew Ng, writing in The Batch, describes the job the same way the postings do: an engineer embedded in a client organization, building and tuning agentic workflows for that client's needs, where communication and business skills carry as much weight as the technical ones. Understanding needs, prioritizing projects, explaining complex technology, and pushing back respectfully when a request is unrealistic. Every one of those is a scored behavior in the discovery track below.
"I did this 25 years before the role had a name. We built India's first fully indigenous bedside monitor, and everything worked in the lab. Then I took it to hospital ERs and watched a patient roll over and flood the ECG leads with noise no test plan had imagined. The field is where the product gets finished. That is the FDE's job."
The market has already decided
Indeed Hiring Lab reports Forward Deployed Engineer postings grew more than 700% year over year. And in 2026 the hyperscalers and the model labs turned the role into a budget line.
Five commitments in a single year turned FDE from a job title into a line item. They split into three ways of funding it:
Amazon ($1B) and Microsoft Frontier ($2.5B) fund internal FDE teams from their own resources.
OpenAI's Deployment Company ($4B, TPG-led syndicate) and Anthropic's services venture ($1.5B, with Blackstone and Goldman Sachs) embed FDEs and, in doing so, bind the customer to their model.
Google Cloud's $750M fund pays system integrators like Accenture and Deloitte to deploy Gemini for customers.
Aggregate framing: Tomasz Tunguz, Theory Ventures. Each commitment independently reported (PYMNTS, TechCrunch, SiliconANGLE), April to July 2026.
"The bottleneck shifted from model capability to deployment."
That is the whole case for this page. When the constraint is deployment and not the model, the scarce resource is people who can walk into an operation and make the AI land. Every dollar above is spent on that human layer. The labs are spending it to embed their people in your operation and standardize it on their stack. The alternative is your own Forward Deployed Engineers, fluent in whatever model wins next year. That is what we build.
Standing up a $1 billion internal Forward Deployed Engineering organization, with the first postings already live.
A 6,000-person program staffed through consulting partnerships. The partners now have to produce FDEs at scale.
A publicly announced Forward Deployed Engineer force embedded with customers.
FDEs embedded with enterprise teams to leave behind patterns the teams own. HP, Intuit, Oracle, State Farm, Thermo Fisher, and Uber signed at launch.
One more thing the postings tell us: of the 10,426 tasks we extracted from FDE job descriptions, only 0.6% can be fully automated. 99% need a human in the loop or are human work outright. Companies are hiring this role precisely because it cannot be replaced by the thing it deploys.
Four capabilities. One continuous engagement.
The whole track runs inside one simulated customer. Your FDEs interview its stakeholders, prototype against its constraints, extend its product, and demo to the people they interviewed. Every stage is anchored on the customer problem, because that is what the role is anchored on.
Learn in GenAI Sandboxes
Your FDEs start inside provisioned environments seeded with the domain: the data shapes, the systems, the vocabulary of the operation they will walk into. Not slideware. Working environments where wrong answers fail visibly.
Practice in the Customer Interview Simulator
Timed conversations with simulated stakeholders who hold hidden facts about the real problem and reveal them only when trust is earned. Pitch too early and the conversation closes. A scored debrief shows what your FDE surfaced, what they missed, and which questions would have gotten there.
Prototype in sandboxes
The problem surfaced in discovery becomes the build target. Your FDEs prototype against masked data inside a sandbox that mirrors the customer constraint they discovered, including the ones IT put there for good reasons.
Extend a working product, then show the customer
The capstone. A reference product is already running, seeded with data and the same gaps real systems have. Your FDE grafts the prototype onto it without breaking what works, then demos the result back to the stakeholder they interviewed. Scored on whether it lands.
From prototype to a product that is already running
Every FDE curriculum on the market stops at the prototype. But the customer does not run prototypes. They run a live product with years of data in it, integrations nobody documented, and users who did not ask for a change. The graft is where FDE pilots die.
So we made the graft the capstone. Your FDE inherits a running reference product, complete with the workarounds and gaps that made the customer problem exist in the first place, and extends its functionality with the prototype they built. It has to work without breaking what already works. Then they demo it to the stakeholder they interviewed in week one, and the demo is scored on whether it answers the problem that stakeholder revealed.
Compose your org's Forward Deployed Engineer job description
Start from the tasks real postings ask for, keep the ones that match your operation, add what is specific to you. The tasks carry their AI classification, so the JD you take away already says what AI runs and what stays with people.
Questions operations leaders ask us
Should we build an internal FDE team or buy vendor FDEs?
Andrew Ng's caution in The Batch is the sharpest version of the tradeoff: vendor FDEs are there to integrate that vendor's product deeply, and in a market where nobody can predict next year's best model, binding your processes to one vendor costs you optionality. In 2026 that stopped being hypothetical. OpenAI ($4B) and Anthropic ($1.5B) each stood up separately funded deployment companies whose engineers embed in your operation and standardize it on their model, which is exactly the switching cost Ng warns about. An internal FDE team keeps the choice of AI stack yours. Both patterns are live in the market, and either way someone has to develop the people. Nuvepro enables the team you choose to build, yours or your partner's, on the same track.
What does an FDE team ship first?
First AI-enabled workflow. Built by your team. 4 weeks. The FDE track anchors on one real workflow from your operation, and the capstone is a working extension of a running product, not a slide deck.
We already have solutions engineers. Is this a new hire or a conversion?
Usually a conversion. Solutions, customer, and implementation engineers already own the customer-facing half of the FDE profile. The enablement track adds the AI-build half. That is a much shorter path than hiring against a role title with 700% posting growth.
Why does the customer problem matter more than the tech?
Because the FDE deliverable is not a model, it is a change in how the customer's work runs. An FDE who can build but cannot surface the real problem ships the wrong thing faster. That is why discovery practice is scored as seriously as the build.
Start with the work, not the org chart.
Run the audit on one operation. See which tasks an FDE team would take on first, and what the first 4 weeks ship.