Reinforcement Learning Infrastructure Engineer at Harmonic

Harmonic is hiring an Reinforcement Learning Infrastructure Engineer to own and optimize its proprietary reinforcement learning training and serving infrastructure. The role focuses on improving throughput and low-latency inference for foundation model workloads, including distributed training, sharded multi-node training, inference routing and scheduling, GPU cluster performance, and custom CUDA or Triton kernels. Candidates need a computer science or related degree, at least two years of relevant industry experience, Python proficiency, and experience with machine learning frameworks and distributed training or GPU cloud infrastructure.