Staff Performance Engineer - ML Training & GPU Kernel Expert

Deepstreamtech; Greater London; On-site; GBP 70,000 - 90,000; Full time

Job summary

Deepstreamtech is seeking a Performance Engineer in Greater London to enhance the performance of their advanced language models. The role demands strong software engineering and machine learning expertise, focusing on optimizing training metrics and developing efficient systems. You will work on CUDA and other tools to drive innovation in natural language processing and collaborate with leading researchers in the field.

Qualifications

• Extremely strong software engineering skills.
• Proficiency in Python and related ML frameworks such as JAX, Pytorch.
• Experience writing kernels for GPUs using CUDA, triton.

Responsibilities

• Optimize performance of advanced language models and systems.
• Improve key model training metrics like training throughput.
• Design high-performant software for training.

Skills

Software engineering skills

Proficiency in Python

Experience with ML frameworks (JAX, Pytorch)

Experience with GPU programming

Experience with distributed training

Familiarity with Transformers

Research paper publication

Job description

Deepstreamtech is seeking a Performance Engineer in Greater London to enhance the performance of their advanced language models. The role demands strong software engineering and machine learning expertise, focusing on optimizing training metrics and developing efficient systems. You will work on CUDA and other tools to drive innovation in natural language processing and collaborate with leading researchers in the field.