Hybrid AI Infrastructure Performance Engineer

CommonAI C.I.C.; Cambridge; Hybrid; GBP 60,000 - 90,000; Full time

Benefits offered by this job

Pension

Professional development

Networking opportunities

Hybrid work

Office near Cambridge station

Job summary

CommonAI CIC in Cambridge is seeking an experienced engineer to design and run reproducible experiments across hardware and software stacks, with a focus on LLM serving systems. You will build measurement tooling, collect traces and profiles, and investigate bottlenecks with a collaborative team.

The role involves collaborating with workload, simulation and uncertainty specialists to calibrate models, define experiment controls, and document repeatable methods.

Qualifications

• Experience designing experiments on real hardware and software.
• Ability to collect traces, profiles and counters for latency, throughput and memory.
• Experience with LLM serving workloads and profiling tools (Nsight, PyTorch Profiler) is useful.

Responsibilities

• Design and run reproducible experiments across hardware or software components and complete LLM serving systems.
• Develop reusable measurement tools and collect traces, profiles and counters for latency and memory behavior.
• Investigate bottlenecks and performance variation across the stack with colleagues.
• Ensure reproducible workload details, software versions, configurations and measurement methods.
• Coordinate measurements with simulation and uncertainty specialists to calibrate models.
• Help define controls for experiments and manage cluster access with hardware teams.
• Establish a reproducible reference experiment and measurement dataset for ongoing work.
• Demonstrate experience in evaluating performance of real computing systems.

Skills

Experiment design

Profiling techniques

LLM serving

Python / scripting

Hardware measurement

Nsight / PyTorch Profiler

System profiling tools

Education

Postgraduate study useful

Tools

Nsight

PyTorch Profiler

System profilers

Job description

CommonAI CIC in Cambridge is seeking an experienced engineer to design and run reproducible experiments across hardware and software stacks, with a focus on LLM serving systems. You will build measurement tooling, collect traces and profiles, and investigate bottlenecks with a collaborative team.

The role involves collaborating with workload, simulation and uncertainty specialists to calibrate models, define experiment controls, and document repeatable methods.