Summary

The performance engineer optimizes AI training and inference workloads using GPU kernels, profiling tools, and distributed training frameworks for materials discovery. This is a hybrid role with three days per week in the office.

Responsibilities

• Analyze architectures for hardware efficiency
• Optimize gpu kernels and collective communications
• Implement quantization, pruning, and distillation
• Profile training jobs to identify bottlenecks
• Build dashboards for compute performance metrics
• Develop scaling tools for growing models