Search
Go
Helix AI Engineer, Training Performance
SAN JOSE · US
$200K–$400K
Full-time
About the Role
Optimize the training performance of 100B+ parameter models on 100k+ GPU clusters for humanoid robots at Figure. The role involves tuning distributed training systems, writing custom CUDA kernels, and evaluating new hardware accelerators. This is a high-impact engineering position focused on scaling AI training infrastructure.
Responsibilities
- Tune distributed training for 100B+ parameter models
- Write custom kernels for optimized performance
- Evaluate and benchmark new accelerators
Requirements
- Experience with distributed training of large-scale models
- Proficiency in GPU programming and CUDA
- Strong understanding of deep learning frameworks like PyTorch
Tech Stack
CUDAGPUDistributed Training
Figure
San Jose · US · 527+ employees
Figure is an AI robotics company focused on developing general-purpose autonomous humanoid robots. The company aims to address labor shortages by deploying robots capable of supporting tasks in manufacturing, logistics, warehousing, and retail.
