Optimize cutting-edge AI models as a Performance Engineer at Cohere. Focus on enhancing training metrics while collaborating in a remote-friendly workplace and contributing to innovations in natural language processing. Cohere is seeking a Performance Engineer to enhance the efficiency of our advanced language models.

In this role, you will leverage robust software engineering skills to design high-performance training software. Collaborating primarily between the Eastern and Central European times, you will tackle performance bottlenecks and develop innovative profiling tools to further our mission. Key Responsibilities:
Design scalable software for optimal training performance
Analyze architectural choices impacting training throughput




Write low-level CUDA and triton kernels for GPU performance
Research and implement solutions on our data infrastructure
Collaborate with leading researchers in AI

Requirements:
Solid software engineering background
Proficient in Python and ML frameworks like JAX and Pytorch
Experience with CUDA and triton for kernel writing
Familiarity with large-scale distributed training
Understanding of autoregressive models like Transformers

Bring your expertise in software engineering and AI to drive impactful innovations at Cohere.

📌 Performance Engineer At Cohere Ai Toronto (Canada)
🏢 Cohere
📍 Canada