AI Infrastructure Performance Engineer

CommonAI C.I.C.; Town of Cambridge; Hybride; AUD 115 000 - 163 000; Plein temps

Avantages offerts par ce poste

Competitive salary

Pension plan

Professional development opportunities

Networking opportunities

Vibrant Cambridge office

Résumé du poste

CommonAI CIC invites a driven engineer to design and run reproducible experiments across hardware and software to illuminate AI infrastructure behavior. You will build measurement tools, collect traces and profiles, and help calibrate simulator models for a growing lab focused on scaling inference.

You will explore bottlenecks, performance variation and energy measurement across components, collaborating with workload, simulation and uncertainty specialists.

Qualifications

• Depth in accelerator profiling, inference runtimes, memory systems or HPC workloads.
• Experience with LLM serving and profiling tools like Nsight or PyTorch Profiler is useful but not required.

Responsabilités

• Design and run reproducible experiments on hardware/software components and complete LLM serving systems.
• Develop reusable measurement tools; collect traces, profiles, and counters for latency, throughput, memory, power and energy.
• Investigate bottlenecks and performance variation across stack layers with colleagues.
• Ensure experimental controls, reproducibility, and calibrated simulator models.
• Collaborate with simulation and uncertainty specialists to select measurements that resolve unknowns.

Connaissances

Accelerator profiling

Inference runtimes

Memory systems

Distributed HPC

Networking

Power measurement

LLM serving

Outils

Nsight

PyTorch Profiler

Description du poste

CommonAI CIC invites a driven engineer to design and run reproducible experiments across hardware and software to illuminate AI infrastructure behavior. You will build measurement tools, collect traces and profiles, and help calibrate simulator models for a growing lab focused on scaling inference.

You will explore bottlenecks, performance variation and energy measurement across components, collaborating with workload, simulation and uncertainty specialists.