Cowboy Space Needs 42 More in Seattle. The Pool Is 750 Nationwide.

On May 11, 2026, Baiju Bhatt's startup (formerly Aetherflux) closed a $275M Series B at a $2B valuation, rebranded as Cowboy Space, and posted 18 roles in a Seattle office it wants to grow to 40-60 people. The job titles read normal: avionics, propulsion, spacecraft design, software. The actual hire is not normal. Cowboy is building a rocket and an 800-GPU orbital data center as a single integrated system, and the engineer who can hold both halves in their head does not exist as a LinkedIn title.

If you are sourcing for this company, or for Starcloud across the street in Redmond, your title-search instincts will fail you in the first hour. Here is what the talent map actually looks like.

The pool is 750 people. Cowboy wants 5-8% of it.

Start with the literal title. Across the entire United States, the count of profiles carrying "Propulsion Engineer" as their current title sits at roughly 721. The top five current employers are Boeing, Northrop Grumman, Blue Origin, Relativity Space, and Venus Aerospace, plus a long tail of US Navy and US Air Force.

721

That number alone should change how you scope this search. Cowboy is not hiring from a generous market. It is hiring from a five-company alumni list, in two zip codes, against four other funded competitors. A "post the JD and wait" approach gets you the same 30 inbound profiles every other recruiter in Kent and Redmond is seeing.

And that is the easy half. Cowboy's actual hard role is the engineer who can treat a 1 MW orbital GPU cluster and a rocket upper stage as one thermal, power, and structural problem. That person almost never has "propulsion engineer" in their title at all.

What Cowboy is actually building (and why the org chart is weird)

The TechCrunch piece on the raise spelled out the spec. Each satellite will mass 20,000 to 25,000 kg and generate 1 MW of onboard power for just under 800 GPUs. The launch vehicle Cowboy is designing in-house lands between Falcon 9 and Starship in capability. NVIDIA confirmed in a May 8 release that Cowboy is one of the launch partners for the Space-1 Vera Rubin Module, with the Rubin GPU delivering up to 25x more AI compute for space-based inferencing than H100.

Translate the spec into roles and you get three pillars that do not normally share an org chart:

1. Propulsion and launch. A new clean-sheet rocket, with Stoke Space and Relativity as the closest founder-DNA matches.
2. Spacecraft bus and structures. A 25-ton satellite is closer to an ISS module than a Starlink bird. Axiom Space and Kuiper alumni are the precedent.
3. Orbital compute. Radiation-hardened thermal, power, and packaging engineering for an 800-GPU cluster running on a moving thermal radiator in LEO. Almost nobody has shipped this.

The named hires so far are the giveaway. Warren Lamont came over from Blue Origin propulsion. Tyler Grinnell came from SpaceX as a launch director. Neither is a satellite person. Both are operators. Cowboy is hiring people who have actually lit hardware, not just designed it.

Why "Launch Director" is the wrong title to clone

Most recruiters, handed the Grinnell hire as a template, will go search LinkedIn for "Launch Director" and "Launch Operations Manager" and come back with 40 names. That misses the point. Grinnell's value is fifteen years of test-stand, range-safety, and ground support equipment (GSE) reps. The people who match that pattern at SpaceX, Blue Origin, and Stoke usually carry titles like "Test Engineer III," "Responsible Engineer, GSE," or "Pad Lead." None of those will surface on a "Launch Director" string match.

This is one of the cases where natural-language sourcing actually earns its keep. With Refolk you can ask "engineers who have run static fires or hot fires at SpaceX, Blue Origin, or Stoke Space in the last four years, based in Washington state" and get back the test-stand population directly, instead of the title-only population.

The Redmond cluster, not Seattle proper

Cowboy's office is technically in Seattle, but the talent gravity is fifteen miles east. SpaceX's Redmond satellite factory builds Starlink. Blue Origin's main campus is in Kent. Stoke Space is in Kent. Aerojet Rocketdyne has a Redmond site. Amazon's Project Kuiper engineering sits in Redmond and Bellevue. Microsoft Azure's HPC and GPU cluster team is in Redmond.

Starcloud, the most direct competitor for Cowboy's exact hire profile, moved itself to Redmond in February 2024 specifically to be close to this cluster. It is not subtle. The Wikipedia entry on the company states the move was made "to be close to space and data center talent at Starlink, Amazon Leo, AWS, and Azure." Then it became the fastest Y Combinator unicorn ever, hitting a $1.1B valuation 17 months after demo day.

Source by prior employer, not current title. The Redmond cluster is four companies wide and one zip code deep. </pull> If you are running Cowboy Space hiring for the Seattle office, your sourcing aerospace engineers Seattle query should be inverted. Do not search Seattle. Search five named companies in Redmond and Kent, filter for people who joined before 2022 (vested and bored), and then filter again for the second skill. ## The second skill: orbital GPU thermal and power This is the bottleneck role and almost nobody is sourcing for it correctly. Look at Starcloud's founding team for the template: - **Adi Oltean.** Chief Engineer. SpaceX Starlink, then Microsoft Azure GPU clusters. This is the literal template hire: a satellite engineer who has also run a hyperscale GPU fleet. - **Philip Johnston.** CEO. McKinsey, then Y Combinator. - **Ezra Feilden.** Airbus Defence and Space. Now look at the NVIDIA Space-1 partner list: Cowboy (Aetherflux), Axiom Space, Kepler Communications, Planet Labs, Sophia Space, and Starcloud. Plus Aethero, which already launched the first space-based NVIDIA Jetson in 2025. Plus Google's Project Suncatcher. Plus Eric Schmidt's Relativity Space, which he acquired specifically to put data centers in orbit. Plus the rumored SpaceX program Musk has alluded to. The actual global population of engineers who have shipped GPU compute into orbit is, generously, 50 to 100 people. The propulsion engineer recruiting market looks tight at 721. The orbital data center talent market is an order of magnitude tighter than that.

refolk prompt: Engineers in Washington state who have worked on both spacecraft thermal or power systems and GPU server hardware, currently at SpaceX, Blue Origin, Kuiper, Azure, or AWS. note: You get a ranked list of the Oltean-shaped hybrids across the Redmond cluster, with current employer, tenure, and a public signal (GitHub, conference talk, patent) for why each one matches both halves. slug: 9r1kn20x0a

### Where the hybrid candidates actually live

Three buckets, in order of yield:

1. **Microsoft Azure HPC and GPU cluster engineers who hold an aerospace degree.** This is the cleanest filter. Azure's GPU fleet team is in Redmond. Aerospace undergrads who took the FAANG offer in 2018-2021 are now five years in, AI-fatigued, and live a 12-minute drive from Cowboy's office.
2. **SpaceX Starlink avionics and power engineers with four-plus years of tenure.** They have shipped flight hardware at volume and have stock that is now liquid. Same for Kuiper, with the wrinkle that Kuiper engineers are easier to recruit because the program is younger and the equity story is less compelling.
3. **NVIDIA, Supermicro, and Dell DGX hardware engineers who keep showing up at AIAA or Small Sat.** Rare, but the people who do this self-select hard. Conference attendee lists are the cheapest signal you can buy.

A title search on LinkedIn will surface none of these reliably. A skills intersection query, asked in English, will. That is the entire pitch for tools like Refolk on a search like this: you describe the two-skill candidate ("Starlink avionics plus GPU cluster experience, Redmond, four-plus years tenure") and get a shortlist, instead of running three separate Boolean strings and trying to mentally intersect them.

## The hiring tells in Cowboy's roadmap

A few details from the research that should shape your candidate pitch, not just your sourcing:

**Cowboy is building its own rocket, not waiting for Starship.** Starcloud-3 is designed to fit Starship's PEZ dispenser, with commercial Starship access expected in 2028-2029. Cowboy decided to vertically integrate launch. This means the propulsion and structures hires need to want to build a new vehicle, not just integrate to one. That maps to Stoke Space and Relativity alumni far better than to Starlink integrators. When pitching, lead with vehicle ownership, not payload.

**Bezos called gigawatt orbital data centers "10+ years out."** Cowboy and Starcloud disagree, publicly, with capital behind the bet. Engineers who want to be on the early-and-right side of that bet are a self-selecting group. The story sells itself if you can get the meeting.

**NVIDIA partnership is a Tier-1 recruiting asset.** "We are deploying Vera Rubin modules in LEO with NVIDIA" is a stronger opener than "we just raised $275M." Use it in the first outreach line, not the third.

## A practical sourcing plan for the next 30 days

If you are running this search, here is the shape that actually works for space data center engineers in this market:

1. **Pull the Refolk index of the five named feeders.** SpaceX Redmond, Blue Origin Kent, Stoke Space, Aerojet Rocketdyne Redmond, Kuiper. Filter to four-plus years tenure. That is your propulsion and bus pool.
2. **Pull the Azure HPC and NVIDIA Redmond pool with an aerospace background.** This is your hybrid pool.
3. **Cross-reference GitHub contributors to NASA cFS, F Prime, and KubeSat.** Open-source flight software is the cheapest second signal for "this person is serious about space," and it surfaces engineers who never updated their LinkedIn after their last job change.
4. **Mine AIAA Propulsion and Energy Forum 2025 attendees and NVIDIA Inception members.** These two lists, intersected, are maybe 200 people nationwide. Buy or scrape them.
5. **Skip "Launch Director" as a title.** Search for test-stand, GSE, and range-safety leads who have actually run a campaign.
6. **Lead outreach with the NVIDIA Space-1 line and equity terms, not the JD.**

Five named feeders. Two skill stacks. One thirty-mile radius. The math is hard but the map is small, which is exactly why a plain-English query against a skill-intersection index will beat ten Boolean strings on LinkedIn every single time on this search.

## FAQ

### How many propulsion engineers exist in the U.S. talent pool?

Based on the Refolk index, roughly 721 U.S. profiles currently carry the literal title "Propulsion Engineer," concentrated at Boeing, Northrop Grumman, Blue Origin, Relativity Space, Venus Aerospace, and the U.S. military. Cowboy Space's plan to hire 40 to 60 Seattle engineers represents 5 to 8% of that national pool, before filtering for the orbital GPU and data center skill intersection that the job actually requires.

### Who is Cowboy Space competing with for the same hires?

Directly: Starcloud (Redmond, $1.1B valuation in 17 months), Aethero, Sophia Space, Kepler Communications, Axiom Space, and Google's Project Suncatcher for the orbital compute side. For the rocket side: SpaceX Redmond, Blue Origin, Stoke Space, Relativity Space (now owned by Eric Schmidt and pivoting toward orbital data centers), and Aerojet Rocketdyne. Microsoft Azure and AWS for the GPU cluster half of the hybrid hire.

### What is the single most valuable hire Cowboy can make right now?

An Adi Oltean clone: someone who has shipped satellite hardware (Starlink, Kuiper, or equivalent) and has also operated a hyperscale GPU cluster (Azure, AWS, or NVIDIA DGX). That intersection is probably 50 to 100 people globally. They will not surface on a title search. They need to be sourced by skill intersection and by alumni overlap between two feeder lists, which is exactly the search Refolk was built to answer.

### Should recruiters wait for layoffs to source these candidates?

No. The Redmond cluster has not had a structural layoff event. The opening is tenure-driven, not layoff-driven: SpaceX Starlink and Blue Origin engineers hired in 2019-2021 are now fully vested, four-plus years in, and seeing Starcloud and Cowboy raise at unicorn valuations next door. The window is open right now. By the time it closes, both Cowboy and Starcloud will be through their next funding round.