Monumental's $32M Khosla Round Is Chasing 2,779 FDEs, Almost None in Texas

On July 15, 2026, Amsterdam's Monumental closed a $32M Series B led by Khosla Ventures to scale its 150-robot bricklaying fleet across Europe and open in Texas, Florida, Virginia, and Arizona. The co-founders (ex-Silk, sold to Palantir in 2016) are pitching this as the first honest attempt to port Palantir's forward deployed engineering playbook to a jobsite. That framing changes the recruiting problem completely: the hire is not a computer vision engineer, and Boolean strings on LinkedIn will miss almost everyone who matters.

Why Monumental's raise is a persona problem, not a keyword problem

Monumental's hire is a hybrid: FDE-shaped, computer vision and embedded fluent, and willing to stand on a slab in Phoenix in August. That combination collapses three separate talent pools into an intersection recruiters routinely underestimate.

The company's Atrium platform runs robots that lay brick and mortar to millimetre precision using sensors, computer vision, and cranes. But the business model is outcome priced. Monumental charges for the finished wall, not the machine. That means every deployed engineer owns a customer relationship, a physical site, and a perception stack simultaneously. Palantir called this role a forward deployed engineer (FDE) starting in the mid-2000s: an engineer embedded with a customer, empowered to ship whatever the deployment needs. Salar al Khafaji and Sebastiaan Visser lived that model at Silk and Palantir. They are explicitly rebuilding it.

Which means the sourcing question is not "who knows OpenCV." It is "who has run a forward deployed loop, can debug a stereo rig, and will fly to Fort Worth on Sunday."

The US FDE pool is 2,779, and the geography is upside down

In Refolk's index of professional profiles, only 2,779 people in the US carry the title "Forward Deployed Engineer" or a close variant. Nearly half of that pool sits in NYC and SF. Monumental's four target states barely register.

Here is the shape of the pool that Monumental is actually fishing in:

2,779

The number itself is not the punchline. The punchline is that FDE culture propagates through Palantir's NYC, DC, and Palo Alto offices, which means anyone willing to work a Phoenix jobsite is self-selecting out of the dominant career geography. That is a much narrower funnel than "2,779 people."

Why the geography matters more than the count

The 18-month benchmark Monumental gave Fortune is "high double to low triple digit" permanently deployed robots across at least two US states. Every robot needs a human within a two-hour drive who can push firmware, debug a mortar viscosity edge case, and pacify a general contractor. You cannot fly a Manhattan FDE to a Tampa slab on demand for eighteen months. The hires have to live there.

The Palantir alumni raid is the obvious move, and the hardest

Roughly 20% of the current US FDE pool still works at Palantir itself, so Monumental's cheapest sourcing motion is a direct alumni raid. It is also the most contested. Every physical-AI startup in the Khosla portfolio is running the same play.

The mechanism to exploit: FDE burnout on multi-year government deployments plus the current physical-AI hype cycle creates a narrow poaching window. The pitch that works is not comp. It is "own a physical outcome, not a slide deck." Ex-Silk and ex-Palantir engineers respond to that phrasing because it names the specific thing they miss.

Companies worth raiding for FDE-shaped talent, per Refolk's index of current employers in the pool:

• Palantir Technologies (still ~20% of the FDE pool, the primary source)
• Modal (infra FDEs who already work embedded with customers)
• Cresta (applied AI deployment engineers with enterprise scars)
• Amp Code (early-stage FDE culture, smaller pool but high signal)
• Built Robotics (closest US cultural analog, argues labor shortage is the bottleneck)

The problem with keyword-first sourcing here is that half of these people do not have "Forward Deployed Engineer" in their headline. They have "Solutions Architect," "Deployment Strategist," "Applied Engineer," or the increasingly common "Field Engineering Lead." A LinkedIn title filter cuts the real addressable pool roughly in half before you have said hello. This is the exact gap Refolk closes: you describe the person in plain English (an ex-Palantir FDE who has shipped a perception stack and is open to Austin or Phoenix) and get a ranked shortlist that ignores the title mismatch.

The CV plus robotics pool is 3x larger, and mostly the wrong hire

The intersection of US engineers with both computer vision and robotics skills is 8,602 people, about 3.1x the FDE pool. That looks like relief until you check where they work.

The top employers in the CV+robotics pool are Meta, Google, Waymo, and Nvidia. Those are perception researchers, not deployment engineers. A staff CV engineer from Meta Reality Labs can absolutely design Monumental's brick-detection network. She will also, statistically, refuse the "stand in the mud" test. Monumental's outcome-priced model requires the engineer to own the customer's finished wall, not the paper. Meta and Google do not train that muscle.

A pure CV engineer from Meta will build a perfect model and quit the second the mortar clogs the nozzle at 6am.

The usable slice of the 8,602 pool is the subset who have already shipped in a startup, worked with customers, and have some hardware scar tissue. That is a subset of a subset. In practice it looks like:

1. Ex-Skydio, ex-Zipline, ex-Cruise engineers who left before the layoffs and want to build again.
2. Ex-Built Robotics, ex-Dusty Robotics people already inside construction who understand the site dynamics.
3. FBR and Fastbrick alumni (Australia, but the bricklaying-specific bench).
4. Amazon Robotics engineers in Boston and Seattle who have run pilot deployments.
5. Ex-Gravis Robotics people if the Zurich raise ($23M in November 2025) produced any early attrition.

The intersection of "FDE-shaped" and "CV/robotics-fluent" and "geography tolerant" is, generously, low hundreds nationwide. Not thousands.

The 6x deployment ramp compresses the timeline to weeks

Monumental built nearly 50 homes in the last three months, up from 8 the quarter before. That 6x quarter-over-quarter ramp means new FDE hires have to be productive in weeks, not months, which pushes sourcing toward alumni networks over cold outreach.

Cold sourced FDEs typically take 60 to 90 days to close and another 90 days to ramp. Referred ex-Silk and ex-Palantir alumni close in 30 days and ramp in 30 because the mental model is already shared. When the fleet is doubling per quarter, that gap is the entire hiring plan.

Practical implications for the recruiter running this search:

• Start every conversation with a Silk or Palantir alumnus by asking who else from their pod is unhappy.
• Weight referrals from Salar and Sebastiaan's personal networks over any external channel.
• Skip Boolean sourcing for the first thirty hires. Use it only for the second wave.
• Prioritize candidates already living in Austin, Dallas, Miami, Tampa, Arlington, or Phoenix. Relocation adds 45 days to the close, which the ramp cannot absorb.

Immigration policy is a hidden tailwind for robotics headcount

Tightening US immigration enforcement is shrinking the informal labor pool that American construction quietly depends on, which raises the ceiling on how many robots each site can absorb, and therefore how many FDEs Monumental can justify hiring.

The mechanism: fewer human bricklayers per site means higher per-robot revenue, which means more headroom to pay senior engineer salaries. The US needs 349,000 net new construction workers in 2026 alone. The UK is short 20,000 bricklayers against a 1.5M-home target, with only 1,990 apprenticeships completed in 2024. Every one of those missing humans is priced into the addressable market Khosla underwrote.

For recruiters, this shifts the pitch. Monumental is not competing with Google comp for a CV engineer. It is competing with the moral clarity of a housing shortfall of 3 to 4 million units in the US. The FDE archetype responds to that framing because it is a real deployment against a real customer problem, which is exactly what Palantir sold them on in 2015 and stopped delivering by 2022.

The Katerra warning every candidate will ask about

Any senior engineer worth hiring will ask about Katerra, which raised over $1B and went bankrupt in 2021 trying to vertically integrate construction. The honest answer is that Monumental is deliberately not Katerra: it charges for the finished wall using its own fleet, rather than owning the entire building.

The comparison to know cold:

The candidate who understands this table is the candidate who converts. The one who does not will churn in month four when the first Texas deployment slips.

What to actually do this quarter

The four moves that will determine whether Monumental hits its 18-month benchmark are all sourcing moves, not engineering ones. Recruiters who treat this as a CV engineer search will lose to those who treat it as an FDE-alumni raid with geography constraints.

1. Map the 2,779 US FDEs by current employer and city, then filter to Sun Belt residents. The addressable subset is low hundreds. Treat every name as a warm lead.
2. Run a parallel search on the 8,602 CV+robotics pool, filtered to people who have shipped at a physical-AI startup, not a lab. That is your second-wave hire.
3. Prioritize referrals from Silk and Palantir alumni over any inbound channel for the first 30 hires. The ramp math does not work otherwise.
4. Skip title filters entirely. Use a plain-English persona query. This is what Refolk is built for: describe the person you actually want, including "willing to relocate to Phoenix," and get names that a Boolean string will miss.

The story here is not that Khosla underwrote a robotics thesis. It is that Khosla underwrote a persona thesis, and the persona is thin, geographically misaligned, and already employed. Recruiters who see that early will build Monumental's US bench. Recruiters who search "computer vision engineer, Austin" will spend eighteen months explaining why the fleet is behind schedule.

FAQ

How many Forward Deployed Engineers are actually in the US?

Refolk's index shows 2,779 people in the US carrying the title Forward Deployed Engineer or a close variant such as Deployment Strategist, Applied Engineer, or Field Engineering Lead. About 48% cluster in NYC and SF, and only one in a top-10 regional sample sits in Monumental's four target states (Austin). This is why a US expansion into Texas, Florida, Virginia, and Arizona is a geography-arbitrage problem, not a keyword problem.

Why can't Monumental just hire computer vision engineers from Meta or Google?

The CV+robotics pool in the US is 8,602 people, about 3.1x the FDE pool, but it is concentrated in perception research roles at Meta, Google, Waymo, and Nvidia. Monumental's outcome-priced model requires the engineer to own a customer relationship, a physical jobsite, and a perception stack at the same time. A research-only CV background does not survive the first mortar clog at 6am on a Phoenix slab.

Who are the best companies to raid for this exact persona?

Per Refolk's index, Palantir Technologies is still the largest single employer of current US FDEs at roughly 20% of the pool, followed by Modal, Cresta, and Amp Code. For the CV plus deployment slice, Built Robotics, Skydio, Zipline, Cruise alumni, and Amazon Robotics deployment engineers are the right target list. Gravis Robotics in Zurich is the direct competitor for the same profile in Europe.

What's the fastest way to source this hybrid persona without keyword filters?

Skip title-based Boolean sourcing entirely for the first thirty hires. Instead, describe the persona in plain English (ex-Palantir or ex-Silk FDE with a computer vision or robotics deployment history, open to Austin, Phoenix, Tampa, or Arlington) and let a sourcing tool rank against that. Refolk was built for exactly this kind of intersection query, where the useful candidates have five different titles and no single Boolean string catches them all.