Member of Technical Staff - Research Engineer

Highlights

• Make large multimodal training runs faster, stable, and numerically reliable.
• Build profiling harnesses that make performance claims trustworthy at scale.
• Turn urgent training bottlenecks into better tools and abstractions.

What they offer

• Equity
• Reasonable travel costs covered

Ideal candidate

A systems-minded engineer who thrives in ambiguous training problems and owns outcomes end-to-end. You love measurements, debugging, and partnering with researchers to make ambitious ideas trainable.

Requirements

• Experience working deeply on large-scale training systems closely with researchers
• Strong PyTorch fluency including comfort reading and modifying low-level training code
• Experience with distributed training concepts such as FSDP, parallelism strategies, activation checkpointing, NCCL
• Hands-on experience improving training throughput, memory footprint, or stability in real runs
• Experience profiling GPU workloads with Nsight Systems, Nsight Compute, torch profiler or custom telemetry
• Understanding of low-precision training and quantization tradeoffs