Founding Engineer

The next generation of science, from new cancer treatments and clean fusion to understanding the origins of the universe, will be built by automated agents running enormous workloads on GPUs. Today, that compute is fragmented across dozens of suppliers. Researchers must hunt for capacity, rebuild environments, move data and integrate a different API for every cloud.

C3 turns this fragmented capacity into one global compute network. A team describes a job once; C3 finds suitable GPUs across suppliers, reproduces the environment, positions the data, routes the work by availability, price and performance, and meters only the time the job is running.

By pooling the world’s GPU capacity behind one interface, C3 aims to make frontier compute accessible to everyone, not only the handful of companies and institutions able to secure it today. Scientific ambition should be limited by ideas, not access to GPUs.

Built during the founder’s PhD in cosmology and machine learning at Cambridge, C3 is live with its first users, has raised a $5M pre-seed, and has cut billed GPU time 15.1x on real research workloads.

C3 is built to serve the rise of automated research. As AI agents generate, launch and iterate on computational work, they need reliable, elastic GPU access without a person managing every provider. C3 provides that infrastructure.

C3 itself operates agent-natively. Internal systems built on the founder’s widely used open-source agent orchestration tools support engineering, research and company building, giving the team unusual leverage from the start.

The hard problems

Operating a GPU cluster inside one data centre is well understood. Making thousands of GPUs across independent clouds behave as one is among the hardest infrastructure problems in computing. Every provider has different APIs, hardware, storage, networking, quotas, lifecycle semantics and failure modes. There is no shared scheduler, control plane or source of truth.

C3 has built a fully custom orchestration engine from first principles to solve this. It coordinates environments, data, execution, logs and billing across providers, making the global GPU market behave as one programmable system. This is the foundation of what C3 is building: the world’s largest virtual GPU cluster. Slurm and Kubernetes stop at one cluster; C3 spans every cloud.

The system must predict as well as react. C3’s ML model forecasts GPU demand so environments and data can be ready before jobs arrive. The execution layer then routes each workload in real time by GPU type, availability, price, performance and reliability. There is no off-the-shelf architecture for this. Every layer demands novel engineering and cutting-edge research.

The role

This is one of C3’s earliest hires and one of the first employees in an ambitious technology startup. You will help shape the engineering, product and company culture from the beginning. As C3 grows, the path is unusually open: this role could develop into senior engineering leadership or broader executive responsibility.

You will build the systems that turn many GPU clouds into one reliable platform. Early work will include:

• Expand the supplier network: build a repeatable integration framework and the plugins that connect each GPU cloud to C3.
• Own reliable execution: improve the Go agent that runs on each machine, including job lifecycle, streamed logs, mounts, failure recovery and metered billing.
• Build intelligent routing: help decide where every workload runs based on GPU type, capacity, price, performance and reliability.
• Bring research into production: turn C3’s demand-forecasting models into systems that prepare environments and data before jobs arrive.

The current stack is a Go CLI and on-machine agents, with a TypeScript control plane on Cloudflare Workers.

Several parts of C3 represent novel scientific research, and we will publish papers from that work. Engineers who are excited by the science can contribute to the research and publications, but neither is a requirement of the role.

Success in the first three months

Supplier coverage from six to twenty-five, about two a week, on a repeatable framework., * Location: at least three days a week in person in Cambridge, working closely with the founder and C3’s first engineer.

• Team: you will join a small founding engineering team and report to the founder.
• Right to work and sponsorship: C3 welcomes applicants of all nationalities. Visa sponsorship may be available for exceptional candidates, subject to UKVI eligibility and C3 obtaining a sponsor licence. Applicants who already have permission to work in the UK, including eligible Graduate-visa holders, are also welcome.
• Process and timing: applications are reviewed on a rolling basis. The process is an introductory call, a technical conversation, then an offer. We move quickly, adapt to the candidate, and are keen for the right person to start as soon as possible.

• 2+ years shipping production code to real users; strong open source work counts. Exceptional early-career engineers are welcome.
• AI leverage. Demonstrable custom agentic workflows are required. Show us systems you built, in or out of work, and how they multiplied your output.

Nice to have

• Integration work suits you: APIs, SDKs and their quirks; making many suppliers behave as one.
• A strong systems background: distributed systems, cloud infrastructure, performance.
• GPU, HPC or research computing experience.
• A PhD or research background in a computational field, with substantial software work., Above all we hire hard working, intelligent people: exceptional performance in an adjacent field can beat directly related experience. Apply through Indeed with your CV and cover letter. The cover letter should start with no more than two sentences describing the most interesting thing you have ever built. This may or may not be work related. Know someone who fits? Pass this on.

Cambridge, UK. Full-time, permanent. £60,000 to £150,000 depending on experience, plus founding equity.