Senior Performance Engineer AI Infrastructure Cambridge (Hybrid)

Pure Resourcing Solutions; Cambridgeshire and Peterborough; Hybrid; GBP 90,000 - 120,000; Full time

Benefits offered by this job

Hybrid work arrangement

Job summary

Pure Resourcing Solutions is sourcing a Senior Performance Engineer for a Cambridge-based AI infrastructure role with a hybrid working pattern. You will extract live metrics from training/inference runs to build models that forecast cost savings and performance gains.

The role requires strong skills in GPU/accelerator optimisation, LLM/DL hardware understanding, and proficiency with profiling and monitoring tools. A degree in CS/math is expected, with Python data tooling as a plus.

Qualifications

• A degree in computer science, mathematics, or something adjacent.
• A track record of building performance models or calculators (Python or spreadsheet-based) that forecast system behavior.
• Hands-on GPU/accelerator code optimisation using CUDA or similar.
• Solid understanding of how LLMs and DL run on real hardware, training vs inference.
• Proficient with profiling tools (Nsight, PyTorch Profiler) and monitoring (Prometheus, Grafana).
• Python for data work with Pandas and NumPy; comfortable with scripting.

Responsibilities

• Sit between research and engineering, pulling real numbers from live training and inference.
• Build models and calculators that translate data into actionable insights.
• Assess if optimisations justify spend and influence hardware decisions.
• Shape procurement and system design for the organisation and members.

Skills

Performance models

GPU/accelerator optimisation

Understanding of LLMs/DL on hardware

Profiling tools experience

Python for data work

Education

Degree in computer science, mathematics, or adjacent

Tools

Nsight

PyTorch Profiler

Prometheus

Grafana

Pandas

NumPy

Job description

Senior Performance Engineer AI Infrastructure Cambridge (Hybrid) £90k-£120k
Nobody quite knows where their compute budget is actually going until someone builds the model that tells them. That's this role.
My client is a Cambridge-based non-profit that exists to stop different parts of the AI world quietly rebuilding the same infrastructure. Rather than a startup, a big enterprise, a government department and a university lab each working out GPU efficiency from scratch, they pool the hard problems and the expertise needed to solve them, so everyone moves faster. It's early days for the organisation but there's serious momentum and serious financial backing behind it already.
They're hiring Performance Engineers at junior and senior level, to sit at the sharp end of that mission.
Day to day
You'd sit between the research and engineering teams, pulling real numbers off live training and inference runs rather than working from theory. From there, the job is building the models and calculators that turn those numbers into an actual answer: will this optimisation help, would a different accelerator be worth the spend, is this architecture change going to pay for itself. Those answers don't stay internal either, they shape what gets bought and how systems get built, for the organisation itself and for everyone else in the membership relying on that judgement.
What you'll bring

• A degree in computer science, mathematics, or something adjacent
• A track record of building performance models or calculators (Python or spreadsheet-based) that actually forecast how a system will behave
• Hands-on GPU/accelerator code optimisation, CUDA or similar
• Genuine understanding of how LLMs and deep learning models run on real hardware, training versus inference, matrix multiplication, KV-caching, that level of detail
• Comfortable in profiling tools like Nsight or PyTorch Profiler, and monitoring stacks like Prometheus and Grafana
• Python for data work, Pandas and NumPy, plus general scripting

Nice to have rather than essential: a postgraduate degree and research background (publications welcome), real depth on inference serving frameworks like vLLM, a stats background, and any open source or research contributions.
Why look twice at this one
It's a rare early seat at something with genuine backing and genuine ambition, where the work you do gets acted on rather than filed away. Competitive salary and pension, hybrid from a Cambridge office, and real exposure to people across the wider AI and academic scene.