Principal ML Performance Engineer

Job Summary

Focused on GPU optimization, the full-time remote Principal ML Performance Engineer will profile and optimize training and inference for advanced machine learning models while managing GPU cluster efficiency and developing benchmarking tools.

Key responsibilities

• Profile and optimize training and inference for structural and generative models, including transformers and geometric deep learning
• Manage GPU cluster efficiency on GCP, overseeing scheduling, utilization, and cost reporting
• Write and tune custom kernels and scale distributed training across multiple nodes

Required qualifications

• Minimum of 6+ years of experience in ML systems, HPC, or performance engineering, with a BS/MS/PhD in CS, EE, or related field
• Deep knowledge of PyTorch internals and hands-on experience with profiling and optimizing performance bottlenecks
• Experience with CUDA and Triton, including proficiency in reading Nsight output
• Strong proficiency in Python and C++ with experience in distributed training at multi-node scale
• Ability to demonstrate improvements made to models through optimization efforts