Head of Hardware

Helios takes AI infrastructure A to Z. We secure the power, acquire the land, and build the data center infrastructure, then either colocate customer hardware or operate the cluster ourselves as a neocloud, selling GPU-as-a-Service at well below hyperscaler pricing. Power is the scarcest input in AI infrastructure, so we solve for it first and build everything else around it.

We closed a seed round led by AlleyCorp, are signing $100M+/year in customer revenue, and are scaling globally. We're hiring a Head of Hardware, reporting directly to the CEO, to own the physical product end to end.

WHY THIS MATTERS

We're optimists about what AI will do: drug discovery in months instead of years, small teams with the leverage of large ones, work that gets augmented long before it gets replaced.

But the buildout has a cost, and today it lands on whoever lives nearest the site. Conventional data centers strain already-tight grids, demand expansions that take years, and consume millions of gallons of water annually in regions that can't spare it. That's the version of this buildout that turns communities against AI entirely.

Helios is the counter-argument. We deploy at the point of generation, so energy producers monetize excess capacity and the grid absorbs no new strain. We cool closed-loop, with no evaporative water consumption. And we build modular, so capacity arrives in months instead of half a decade.

This role is where that becomes real or doesn't. The module design, thermal architecture, integration line, and delivery schedule are the difference between the thesis and a slide about the thesis.

STAYING AT THE FRONT OF COMPUTE

Accelerated computing re-architects itself every twelve to eighteen months. Densities climb, cooling requirements shift, and the enclosure that was right for one GPU generation is wrong for the next. Companies that treat their physical design as finished get one good deployment cycle, then spend years catching up.

We intend to be the opposite, and we're building a community around that: engineers, energy operators, OEMs, and researchers focused on what's next in compute, not what shipped last year.

Our current module is v1. You'd own v2, v3, and everything after: architecting each generation around the best hardware available, so a new GPU platform or cooling approach lands in a module designed for it rather than retrofitted to it.

THE ROLE

You own our physical product from design through energized, running capacity at site:

Module design and build. Own the design of our modular GPU data centers: mechanical, thermal, power distribution, and compute integration into the enclosure.

The platform roadmap. Own v2, v3, and beyond, architecting each module generation around the newest compute and cooling technology available.

The integration line. Run warehouse integration: build, burn-in, QA, and ship repeatable units, not one-off builds.

Thermal architecture. Own direct liquid cooling end to end: CDUs, facility loops, rack manifolds, closed-loop rejection, at NVL72-class density and beyond.

Power distribution. Own everything downstream of the interconnect: high-density distribution, switchgear, and the medium-voltage systems that feed the modules.

Supplier and OEM program management. Hold OEMs to spec and schedule, manage long-lead items, qualify second sources, and keep the BOM from constraining deployment velocity.

Site delivery. Get modules delivered, installed, powered, cooled, and accepted at each site, working with EPC and O&M partners.

The team and the network. Hire and lead hardware engineering, integration technicians, and deployment staff, with license to build from day one. You'll also be a senior technical voice externally with vendors, partners, and the broader infrastructure community.

The loop with GTM and software. Feed real capability, cost, and lead-time data into what sales can commit to, and make sure what gets sold is deliverable on the timeline it was sold on.

This is a player-coach seat. You're still in the design reviews, on the shop floor, and on site, while building the team that eventually takes those things off your plate.

HOW WE WORK

Our culture mirrors our infrastructure: decentralized and built for speed.

Pragmatism over perfection. A modular, good-enough unit deployed today beats a perfect data center stuck in a five-year permitting phase. We build where the power already is.

Own the outcome. No passengers, only drivers. When you see a gap, you fill it. Anyone can claim an unowned problem off our public gap board, regardless of title.

Reduce decision latency. The greatest risk is moving too slowly. If a call is reversible and you're 70% confident, make it. Tools and parts under $500 need no approval.

Build, break, iterate. Every Friday we demo what actually shipped: a live deployment, a hardware assembly, a permit secured. If it isn't demo-able, it isn't shipped. Post-mortems are blameless and aimed at the process, never the person.

Win-win infrastructure. Our success is tied to the energy producer's monetization and the AI company's latency and cost. If they don't thrive, we don't exist.

MUST-HAVES

Has personally delivered multi-MW GPU deployments, not just designed them. You've owned capacity that went live, on a schedule, with the problems that come with it.

Direct liquid cooling fluency: CDUs, facility loops, rack manifolds, and the failure modes that matter at NVL72 density.

High-density power distribution depth, plus enough medium-voltage and substation literacy to be credible across the table from a power plant.

Modular, prefab, or systems-integration background. You've shipped repeatable units out of a shop, not just built one prototype well.

Strong supplier program management. You hold OEMs to spec and schedule and manage long-lead items without surprises.

Player-coach who builds teams. High ownership, comfortable in ambiguity, still hands-on.

Genuinely follows this field. You have opinions about what the next generation of enclosure needs to look like, not just the current one.

STRONG PLUSES

Miner-to-HPC conversion or crypto-to-AI retrofit experience.

Neocloud or GPU cloud operating experience.

Off-grid, behind-the-meter, or edge/modular deployment background.

Experience standing up an integration facility from scratch.

Comfort with the regulatory and utility interconnection side of siting.

A record of contributing publicly: open designs, standards bodies, conference work, or industry groups.

WHY HELIOS

You own the physical product outright, reporting to the CEO, at a company where the hardware is the differentiation, not a cost center supporting someone else's software. You set the platform roadmap, build the team, and build the technical network around it. Competitive salary plus meaningful founding-team equity. And a scarce, durable problem: power-constrained AI compute is one of the defining infrastructure bottlenecks of the decade.

COMPENSATION

Competitive base + meaningful founding-team equity.