Flourish's $500M Bezos Round Is Chasing ~800 People LinkedIn Can't See

On June 5, 2026, Flourish came out of stealth with a $500M round at a $2.5B valuation, backed by Jeff Bezos (nearly $100M of it), Lux Capital, GV, and Catalio. The plan is Cortex AI: foundation models built around biological cortical columns, targeting a draw of 20 to 50 watts versus the 600-plus a server GPU pulls today. If you run sourcing at a frontier lab, a chip startup, or a neuro-adjacent team, the next ninety days are when the global hybrid pool gets locked up.

That pool is small. Senior researchers who can credibly do both wet-lab connectomics or computational neuroscience and modern deep learning at scale number maybe 800 people worldwide. Most of them are not on LinkedIn in any useful way. They are on arXiv, OpenReview, GitHub, and university lab pages.

What Flourish actually has to hire

Read the WIRED reporting and the FinSMEs writeup together and the staffing plan becomes obvious. Flourish moved into a West SoHo building with a built-in data center and had roughly two dozen neuroscientists and AI researchers in seats by the end of March. They intend to stand up an in-house neuroscience lab with high-resolution electron microscopes, build an AI memory management system, and scale hardware partnerships in parallel.

That is three hiring tracks under one roof:

1. Wet-lab connectomics. Electron microscopy operators, tissue prep, image segmentation engineers. Pull from HHMI Janelia FlyEM, Allen Institute, the MICrONS consortium, Lichtman lab at Harvard, Seung lab at Princeton.
2. Frontier ML engineers. Transformer-era research engineers who can also read a neuroscience paper without flinching. Pull from DeepMind's NeuroAI group, Meta Reality Labs, Numenta veterans, Cortical Labs.
3. The bridge layer. The ~800 hybrids. These are the load-bearing hires. They are also what Groq, Cerebras, Etched, and every hyperscaler with a neuro-inspired sidecar project will fight for.

The wet-lab and ML populations rarely overlap on an org chart. The bridge layer is what makes the whole thing function, and it is the layer LinkedIn surfaces worst.

Why LinkedIn misses this cohort

Try it. Search "computational neuroscientist" with a deep-learning filter on LinkedIn Recruiter. You will get a few hundred profiles globally, most of whom are either students, adjacent (pure ML with one undergrad neuro class), or already inside DeepMind or Meta and locked down. The senior hybrids list themselves as "Research Scientist" or "Postdoctoral Fellow" with a lab name you have to recognize to value.

The signal lives elsewhere:

• arXiv and bioRxiv authorship. The October 2024 Nature paper "The fly connectome reveals a path to the effectome" lists Pospisil, Dorkenwald, Matsliah, Sterling, Schlegel, Yu, McKellar, Costa, Eichler, Jefferis, Murthy, and Pillow. Twelve names. That alone is a shortlist.
• OpenReview profiles. The NeurIPS NeuroAI Workshop author list is effectively a searchable index of the cohort Flourish needs. Its own description, "researchers from neuroscience and AI exploring intersections that promise novel insights into biological neuronal function and the development of computationally less intensive artificial systems," is the job spec.
• GitHub orgs. seung-lab, CAVEconnectome, flyconnectome, google-research/connectomics, MouseLand. Commit history tells you who actually ships versus who shows up on papers as author #14.
• Lab webpages. Princeton (Seung, Murthy, Pillow), Janelia, Allen, Harvard. These rosters change quietly.

~800

If your sourcing motion starts with a Boolean string on LinkedIn, you have already lost this one. This is exactly the gap we built Refolk for: you describe the person in plain English ("co-authored connectomics papers in the last 3 years, also commits to PyTorch-based segmentation tools, US-based") and get a ranked shortlist drawn from GitHub, LinkedIn, and the open web together, not just whichever surface a profile happens to be best maintained on.

The CTRL-labs diaspora is the single highest-yield network

Thomas Reardon co-founded Flourish in 2024 with Rob Williams. Reardon's history matters operationally, not just narratively. He built Internet Explorer at Microsoft in 1994, co-founded CTRL-labs (Meta acquired it in 2019 for somewhere between $500M and $1B), then led neuromotor interface projects at Meta Reality Labs.

Founders hire from the people they have managed. The CTRL-labs alumni network and Reardon's Reality Labs reports are the first ring of recruiting at Flourish. They are also where competitors should be mapping right now, before Flourish closes the network off through offers, equity, and friendship debt.

Two adviser names from the WIRED reporting are worth treating as graph anchors: Greg Wayne (DeepMind) and Jacob Vogelstein. Both have public co-author graphs going back a decade. Walk one degree out from each and you have a list. Walk two degrees and you have the addressable market.

Founders hire from the people they have managed. The CTRL-labs network is the first ring. Map it before Flourish closes it.

The alt-architecture summer

Flourish is not the only $400M-plus alt-architecture raise this quarter. Generalist AI closed $400M at a $2B valuation the same week to build general intelligence for robotics. Groq, Cerebras, and Etched are all hiring against the same hybrid skill set Flourish needs, framed as "AI architecture" rather than "neuro-AI" but drawing from overlapping schools.

Indeed currently lists 87 Computational Neuroscience AI jobs in the US. Against a senior hybrid supply of roughly 800 globally, that is a brutal ratio before you add the unposted roles inside DeepMind, Anthropic, and Meta's FAIR successor teams. Compensation comps for this exact profile do not exist yet because the role barely existed in industry until 2024. Expect that to change within six months. Until it does, first movers can sign people at academic-adjacent prices.

Where the supply actually sits

Working from the research, the obvious nodes:

• NYC area: Basis Research Institute, which already requires "a PhD in statistics, programming languages, machine learning, computational neuroscience, cognitive science, or physics." Flourish will poach here directly, and so should you.
• Princeton: Seung lab, Pillow lab, Murthy lab. Connectomics + theory + behavior, all in one zip code.
• Janelia / HHMI: the FlyEM and FlyWire crowd. Many are postdocs three to five years in, looking at academic markets that have not recovered.
• Allen Institute (Seattle): the MICrONS consortium veterans.
• Google connectomics (Viren Jain's group): co-author on "Next-generation AI for connectomics" in Nature Methods 2024. The bench is deep and underpoached.
• Veterans of IBM TrueNorth, Intel Loihi, SpiNNaker. Neuromorphic computing has not displaced GPU-driven technology in commercial settings. That history makes the cohort skeptical and underpriced. Your pitch needs to be different from the one OpenAI uses. Lead with the connectomics-as-blueprint angle, not the energy-efficiency one (they have heard that pitch for fifteen years).

A 30-day plan if you compete with Flourish for any of this pool

Day 1 to 7: build the graph.

• Pull the author list from the NeurIPS 2024 NeuroAI Workshop on OpenReview.
• Pull authors of the fly connectome Nature 2024 paper and the Nature Methods 2024 connectomics review.
• Pull contributors to seung-lab, CAVEconnectome, flyconnectome, google-research/connectomics, MouseLand. Filter for commits in the last 18 months.
• Cross-reference against CTRL-labs alumni and Meta Reality Labs neuromotor interface reports.

Day 7 to 14: enrich.

• Find current affiliation, location, and reachable contact for each name. For roughly half the pool, this means a lab page email and not a LinkedIn InMail. This is the step that breaks most ATS-driven recruiting orgs, because their tools are LinkedIn-shaped. Using Refolk here collapses what is usually a week of manual cross-referencing into a single query: ask in plain English, get the cross-source profile back.

Day 14 to 30: outreach with the right pitch.

• Do not lead with comp. This cohort took academic discounts on purpose.
• Do lead with the science: connectomics-as-architecture, cortical columns as inductive bias, the specific paper of theirs you read.
• Mention the lab build (electron microscopes in the same building as the GPUs) if that's true for your org. For Flourish it is. For most competitors it is not, and that is a defensible recruiting differentiator the other way: "we won't make you also be a microscopy tech."

What the pool looks like in 12 months

Two scenarios.

If Cortex AI shows even a directional result at the 20 to 50 watt target, the value of a senior neuro-AI hybrid re-prices upward by two to three times, and every hyperscaler stands up a parallel team. The 800-person pool becomes the most contested hiring cohort since the post-Transformer ML rush of 2018.

If it doesn't, the pool stays cheap and the smart move is to hire from it now anyway, because the connectomics datasets coming out of MICrONS, FlyWire, and the Wellcome mouse-brain project are going to feed ML research regardless of whether Flourish's specific architecture wins. Either way, the people who can read both kinds of paper become more valuable, not less.

The mistake is assuming you have a year. Flourish closed a $500M round in about five weeks of conversations after late April. The same speed will apply to hiring. By Labor Day, the most reachable third of the pool will be off the market or in late-stage conversations with one of three or four labs. The other two-thirds will still be reachable, but only by teams that already did the graph work in June and July.

FAQ

Why is the addressable pool only ~800 people?

The intersection of senior computational neuroscience and frontier deep learning is genuinely narrow. Most neuroscience PhDs do not train in modern ML, and most ML researchers do not have wet-lab or biological-systems exposure. The hybrid is usually a postdoc or research scientist with a neuro PhD who joined a lab like DeepMind's, Seung's, Janelia's, or one of the connectomics consortia in the last five years. Add in veterans of IBM TrueNorth, Intel Loihi, SpiNNaker, and Numenta, and you land at roughly 800 globally. That estimate matches what shows up if you walk the NeurIPS NeuroAI workshop, the connectomics paper author graphs, and the relevant GitHub orgs.

Won't Flourish just poach from Meta Reality Labs and DeepMind?

For the first two rings, yes. Reardon's network at Meta Reality Labs and CTRL-labs is the obvious first draw, and DeepMind's NeuroAI group (where adviser Greg Wayne sits) is the obvious second. But Flourish needs to hire dozens of people, not five, and the wet-lab connectomics roles in particular don't exist inside DeepMind or Meta in the volume Flourish needs. That forces them out into the academic networks, which is exactly where competing labs can still get in front of the same candidates.

What's the right opening message to a neuro-AI researcher?

Reference a specific paper of theirs from the last 24 months and what your team is doing that connects to it. Skip comp on the first message. This cohort selected into academia knowing the pay tradeoff; leading with money signals you don't understand them. The reply rates we see on cold outreach to academic-adjacent researchers improve substantially when the first sentence is technical and specific, not transactional.

How does Refolk help with a pool that lives on arXiv and GitHub?

Refolk indexes GitHub, LinkedIn, and the open web (including lab pages, arXiv author profiles, and OpenReview) and lets you query in plain English. For a search like "connectomics paper co-authors since 2022 who also commit to PyTorch repos, US-based or relocatable," you get a single ranked list with the cross-source context attached, instead of building it by hand across five tools. That is the difference between a sourcing motion that finishes in a week and one that takes a quarter, which is roughly the window you have before Flourish locks down the most reachable third of the pool.