Lead Machine Learning Research Engineer

Blankstate is a London-based AI infrastructure company. We have pioneered a new SOTA model topology that independently measures interactions. Our architecture is federated and non-retentive.

We are looking for a Lead Machine Learning Engineer who owns the engineering bridge experimentation between deep model research and the production ecosystem we ship. You would run the seam directly: translating research findings into engineering work, regressions into research questions, and production observations back into experimental probes accompanied with our different units.

This is a senior engineering seat with potential mentorship. You work directly with the Chief Data & Analytics Office and the Global Delivery team, with technical autonomy and a direct line into the science. The design space is open, the constraint set is novel, and the architectural choices you make in your first year become the ecosystem the team builds on.

• Six years or more building and shipping ML systems that customers used and that you were responsible for end-to-end, from offline metric to user-visible behaviour.
• Strong on at least one of: federated or edge ML systems; evaluation and orchestration; simulation-based testing and benchmarking; compiler-style configuration systems, or anything in that family.
• An instinct for making things simpler alongside making them more capable. We hire people who reduce complexity, not just add to it.
• Comfortable in both owning a multi-surface engineering programme and deepen one component.
• The kind of senior who unblocks others by default.
• Fluent in PyTorch and modern serving infrastructure; comfortable engaging with the front-end team on Rust and React adjacent product work even if not writing it.

• Privacy-preserving ML in production: differential privacy, federated averaging, secure aggregation, on-device inference at scale.
• Evaluating modern models under realistic constraints: tool use, multi-step reasoning, memory, parallel execution.
• Open-source contributions to evaluation harnesses, simulation frameworks.
• A published research idea you shipped as customer-visible product behaviour.

• Base salary of £95,000 – £140,000 annually (dependent on experience).
• Meaningful share-pool participation in a UK-based deeptech company.
• Pension and standard UK benefits.
• A conference budget that funds at least one tier-1 venue per year as a co-author or first-author submission.
• Full access to Blankstate's research compute. No notebook-only roles.
• A direct line to the cofounder, the operational research-leadership seat for the next two years, and a clear path to grow alongside the Scientific Director once that seat is filled.
• Flexible working from our London Research Unit next to Hyde Park, with a culture that rewards depth and rigour over performative output.