ML Engineer - HPC Platforms, Low-Latency Inference

Dex; Greater London; On-site; GBP 90,000 - 140,000; Full time

Job summary

Dex is seeking a highly skilled ML/HPC engineer to join a leading quantitative research environment. You will evaluate and integrate cutting-edge ML accelerators and HPC technologies, optimizing inference to meet extreme latency and throughput targets.

You will benchmark distributed training, understand networking/storage trade-offs, and collaborate with internal experts and external vendors to drive platform improvements from concept to robust systems.

Qualifications

• Postgraduate degree in ML or related quantitative field, or equivalent commercial experience building ML systems at scale.
• Strong Python, PyTorch and NumPy programming skills.
• Deep expertise in at least one: optimization methods, modern ML techniques, high-performance computing, profiling, or model inference.
• Experience evaluating and integrating novel ML/HPC technologies with external experts and vendors.

Responsibilities

• Evaluate and integrate cutting-edge ML accelerators and HPC technologies.
• Optimize model inference for extreme latency and throughput requirements.
• Benchmark distributed training setups, understanding networking and storage trade-offs.
• Engage directly with leading technology vendors, influencing their product roadmaps.
• Design and implement solutions, from rapid proofs-of-concept to robust system architectures.

Skills

Python

PyTorch

NumPy

Education

Postgraduate degree in ML or related field

Equivalent commercial ML systems experience

Job description

Dex is seeking a highly skilled ML/HPC engineer to join a leading quantitative research environment. You will evaluate and integrate cutting-edge ML accelerators and HPC technologies, optimizing inference to meet extreme latency and throughput targets.

You will benchmark distributed training, understand networking/storage trade-offs, and collaborate with internal experts and external vendors to drive platform improvements from concept to robust systems.