Research Engineer, ML Infrastructure

Make our models performant, resource efficient and reliable at scale by developing the core frameworks for model training and evaluation, in close partnerships with fellow researchers and engineers.

• Build our training stack across model, layer, and kernel levels; optimize workloads through parallelism, quantization, and custom kernels.
• Profile end-to-end training runs on large GPU clusters; eliminate bottlenecks and failures; monitor throughput, utilization, and uptime.
• Ensure new model architectures and training recipes scale efficiently, from early experiments to frontier-scale runs.