Join Cohere as a Performance Engineer, optimizing AI language models for efficiency. Work remotely or from key offices while enhancing model performance through cutting-edge engineering.As a member of the Pre-Training team, you will focus on optimizing training throughput and improving model metrics. Your expertise in software engineering, CUDA, and machine learning frameworks like Python and JAX will be essential. The role involves identifying bottlenecks and developing training tools to support Cohere's mission in AI.Key Responsibilities:
Design high-performance software for model training
Analyze architectural designs affecting throughput and quality
Write CUDA and Triton kernels for maximum accelerator efficiency
Implement research ideas on supercompute and data infrastructure
Collaborate with leading researchers in natural language processingRequirements:
Robust software engineering skills and Python proficiency
Experience with CUDA and GPU kernel writing
Familiarity with distributed training strategies
Knowledge of autoregressive sequence models
Background in machine learning frameworks like Pytorch and JAXBring your passion for AI optimization and engineering to enhance Cohere’s capabilities.#J-18808-Ljbffr