Join Cohere as a Performance Optimization Engineer and drive the efficiency of our language training models. This role combines advanced software engineering with cutting-edge machine learning in a remote-friendly setting.
As part of our agile Pre-Training team, you'll work on maximizing training throughput and optimizing model architecture. You will utilize your strong background in software engineering and development to create robust systems and identify performance bottlenecks. Your work will directly contribute to enhancing our capabilities in natural language understanding.
Key Responsibilities:
• Develop high-performance training software solutions
• Evaluate architectural modifications for training enhancements
• Write effective CUDA and triton kernels for GPU optimization
• Conduct research on supercompute and data infrastructure
• Collaborate with top-level AI researchers
Requirements:
• Expertise in software engineering principles
• Proficiency with Python and ML frameworks like PyTorch
• Experience in GPU kernel development using CUDA
• Familiarity with distributed training techniques
• Knowledge of Transformers and autoregressive models
Join a team dedicated to AI excellence at Cohere, where your skills will significantly influence our growth.
#J-18808-Ljbffr