Head of Compute

Prime Intellect•San Francisco, CA

9h•Hybrid

About The Position

Prime Intellect is building the open superintelligence stack, encompassing frontier agentic models and the infrastructure to create, train, and deploy them. The company aggregates and orchestrates global compute into a single control plane, paired with a full RL post-training stack including environments, secure sandboxes, verifiable evals, and an async RL trainer. This infrastructure enables researchers, startups, and enterprises to conduct end-to-end reinforcement learning at frontier scale, adapting models to real tools, workflows, and deployment contexts. Prime Intellect recently secured $15 million in funding, bringing their total to $20 million, with investments led by Founders Fund and Menlo Ventures, alongside prominent angels. The Head of Compute will be responsible for all aspects of compute at Prime Intellect, including sourcing, economics, contracting, and the strategic positioning of GPU supply essential for training, serving, and selling. This role is considered highly consequential within the company and the industry, as compute is both the product and a critical constraint for the open AI ecosystem. The individual will play a pivotal role in ensuring the open ecosystem has access to scarce GPUs. Decisions made in this role will influence the scaling of neoclouds, the development of global compute hubs, the accessibility of hardware generations, and the economic viability of training and serving open models. The Head of Compute will collaborate with research and engineering teams to co-design the compute layer of the open model ecosystem, determining training strategies, hardware usage, locations, and cost structures. The position involves direct engagement with leadership teams and founders of neoclouds, as well as research and engineering teams, to establish commercial relationships, secure early access to advanced accelerators, and build a robust financial and operational architecture to create a lasting advantage for Prime Intellect. The ideal candidate must be adept at modeling unit economics, understanding research requirements, and negotiating high-value capacity agreements.

Requirements

Strong business and financial instincts — you think natively in unit economics, margin structure, and capital allocation, and you can model the long-term P&L consequences of complex supply decisions
Deep fluency in the global AI compute market: you know the providers, the hardware generations and their real tradeoffs, the pricing dynamics, and where the market is going over the next 12–24 months
Enough technical understanding to be dangerous — you can push back on a vendor's spec sheet, read a cluster topology diagram, understand why two nominally identical clusters deliver different goodput, and have a real conversation with researchers about what their workloads need
Serious commercial chops: you've negotiated and closed contracts at meaningful scale, know how to find and use leverage, and understand how deal structure drives downstream economics
Comfortable operating at the intersection of finance, commercial, product, and engineering — and translating fluently between all of them
High ownership: you see gaps and build the fix before anyone asks
AI-native in how you work: you use LLMs, automation, and programmatic tools to move faster

Responsibilities

Own the economics of compute end-to-end: the unit economics of every contract, the margin architecture across training and inference products, the long-term P&L consequences of today's supply bets
Partner with Finance and leadership on capital strategy — how much to commit, to whom, for how long, on which hardware, with what balance sheet exposure
Build the frameworks that turn supply decisions into clear financial outcomes, and that let us make multi-hundred-million-dollar bets with conviction under uncertainty
Shape the commercial architecture of the open compute ecosystem: how committed capacity, spot markets, credit structures, and partner economics fit together
Own end-to-end procurement of GPU capacity globally — across hyperscalers, tier-one neoclouds, regional operators, and emerging providers in North America, Europe, the Middle East, and Asia
Negotiate and close reserved capacity agreements, spot and burst arrangements, MSAs, DPAs, and order forms at nine- and ten-figure scale
Secure early access to the latest generations of accelerators (B200, GB200, and what comes next) — in the quantities we need, before our competitors
Build and maintain the senior relationships that make Prime Intellect the partner of choice for providers deciding where to allocate scarce capacity
Work closely with our research team to translate training roadmaps, RL workloads, and open model ambitions into concrete compute requirements — and back the other way, to surface what's possible given the supply we can secure
Partner with Engineering on acceptance testing, goodput validation, and the technical qualification of new providers and hardware
Sit at the table where the biggest calls get made: which open models we train, which customers we serve, which bets are worth the capital
Track pricing, availability, and provider dynamics continuously across every major global market
Serve as the internal source of truth on the compute market — who's credible, who's mispriced, where supply is about to tighten, which providers will still exist in 18 months, where the next wave of capacity is coming online
Advise leadership on the strategic bets that define the company: which accelerators, which providers, which geographies, which contract structures, which moments to lean in hard