Free cookie consent management tool by TermsFeed Generator Founding Performance Engineer – Frontier AI Lab, London | Axioma Search
Image
Image
Bg

Founding Performance Engineer

About

Start from the physics. What’s the best performance an AI model can reach, and what would the hardware and software need to look like to get there? This role owns that question, from the application down to silicon.

This is a VC-backed frontier lab building end-to-end chip design systems in London. The founders have built high-performance chips, and the small team has already built internal systems that improve themselves. The next step is scaling them, with a larger raise expected in early 2027.

The work spans low-level optimisation, performance modelling and making architectural decisions around latency, throughput and hardware constraints.

What you’ll do

  • Own end-to-end performance from application code to hardware: latency, throughput, and where each one is lost
  • Write and optimise low-level code in C++ or Rust, close to the metal
  • Build performance models that inform architecture decisions, then prove them against real code
  • Profile workloads down to memory hierarchy, pipelining and branch behaviour
  • Make hardware-software trade-offs alongside the chip design and AI engineers
  • Decide how far AI-written code can be trusted on performance-critical paths, and build the checks that tell you

What you’ll need

  • A track record of writing highly performant code in production – inference serving, accelerators, HPC, low-latency systems, embedded, defence or aerospace
  • Real hardware awareness: ISAs, caches, memory bandwidth, branch prediction, accelerator architectures
  • Strong C++ or Rust, though the team is open on language if the systems depth is there
  • Ability to reason from first principles about performance limits, and explain it simply
  • A considered view on building and maintaining software when AI writes most of the code
  • Comfort in a tiny team with loose specs and fast decisions

Optional Bonus

  • CUDA, Triton or custom kernel work, or inference optimisation at scale
  • Performance modelling or architecture exploration for chips or accelerators
  • Exposure to RTL, FPGA or silicon bring-up

Shortlisted candidates will be contacted within 48 hours.

Back to job listings
  • Location London
  • Salary / Compensation Up to £120k + equity
  • Work Setup hybrid
  • Sectors Deep Tech, Frontier AI / Foundation Models
  • Skills C++, Rust, Performance Engineering, Computer Architecture, CUDA, Hardware-Software Co-Design
Image

Role Contact

Alex Jouatte

Bg

Didn't find the right role?

Send us your CV.