GPU Performance Research Engineer for Large-Scale LLMs

Mistral Ai; Paris; On-site; EUR 90,000 - 140,000; Full time

Benefits offered by this job

Healthcare coverage

Parental leave

Retirement plans

Relocation support

Wellness programs

Meal and transportation allowances

Job summary

Mistral AI is seeking a Research Engineer, GPU performance, to optimize GPU utilization for frontier models. You will write low-level code and collaborate with researchers to improve inference efficiency at scale.

You will also integrate optimisations into an MLOps framework, profile large-scale training and inference, and build accurate hardware usage estimates across multiple GPU architectures.

Qualifications

• Strong experience in writing low-level GPU kernels and optimising memory layouts.
• Experience with large-scale distributed GPU clusters and orchestration.
• Solid understanding of generative AI concepts and model workflows.
• Good testing, CI/CD and code-review practices.

Responsibilities

• Develop low-level code to maximise GPU capacity, from kernels to memory layout and communication.
• Rethink model architecture for efficient inference in collaboration with researchers.
• Integrate efficient code into a high-level MLOps framework for scalable training and deployment.
• Profile and benchmark training and inference across thousands of GPUs.
• Estimate LLM serving performance across diverse GPUs.

Skills

GPU kernel development

Performance profiling

Parallel computing

Software design

Education

PhD or MSc in Computer Science/EE

Tools

CUDA

MPI

PyTorch

Job description

Mistral AI is seeking a Research Engineer, GPU performance, to optimize GPU utilization for frontier models. You will write low-level code and collaborate with researchers to improve inference efficiency at scale.

You will also integrate optimisations into an MLOps framework, profile large-scale training and inference, and build accurate hardware usage estimates across multiple GPU architectures.