SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Mistral is a full-stack AI company providing frontier models, developer tools, applications, and compute infrastructure. They partner with enterprises across finance, manufacturing, defense, healthcare, and the public sector to build customized AI systems.
As a Research Engineer focused on GPU Performance, you will own the end-to-end performance optimization of Mistral's large language models. Your work directly impacts how efficiently frontier models are trained and served at scale.
Key Responsibilities:
- Develop low-level code that maximizes GPU utilization and capacity, including custom kernels, memory-layout optimization, and communication protocols
- Rethink generative model architectures in collaboration with Research Scientists to improve efficiency for inference and training
- Integrate low-level optimizations into high-level MLOps frameworks to ensure benefits scale across all training runs and deployments
- Profile, benchmark, and extract performance gains from large-scale training and inference across thousands of GPUs
- Model LLM execution on various GPU types with high fidelity to assess hardware usage and build serving performance estimates
You will work at the intersection of systems engineering and AI research, directly collaborating with Research Scientists and working across distributed teams in Europe, North America, Asia, and the Middle East.
Requirements:
- High technical competence in writing custom low-level kernels and pushing GPUs to their limits
- High expertise in distributed computation infrastructure of current-generation GPU clusters
- Overall understanding of generative AI; knowledge of or interest in fine-tuning and applying language models
- Strong software-design instincts including testing, code review, and CI/CD practices
- Self-starter mentality with low-ego, collaborative approach
