M
Posted 5 hours ago
Member of Technical Staff - Efficient ML
MoonlakeMember of Technical Staff - Efficient ML
Requirements
Experience with FSDP/ZeRO/tensor+pipeline parallel, NCCL tuning, Nsight profiling, Triton/CUDA kernels, Inference optimization, Quantization (GPTQ/AWQ), SLURM/K8s
Skills
PyTorchCUDATritonMachine LearningGPU Optimization
About the role
Responsibilities
- Improve training efficiency through dataloaders, fusion, activation rematerialization, and gradient checkpointing
- Implement and tune FSDP, ZeRO, tensor parallelism, and pipeline parallelism using NCCL
- Optimize GPU and kernel performance using Nsight profiling, Triton, and CUDA kernels
- Develop Flash-attention-style speedups, sequence packing, and KV-cache optimizations
- Drive inference optimization including low-latency serving, continuous batching, and speculative decoding
- Implement quantization (GPTQ/AWQ), distillation, and pruning techniques
- Manage infrastructure and reliability using SLURM/K8s for multi-node jobs and ensuring checkpoint hygiene
- Maintain determinism, environment pinning, and GPU failure handling
About the Company
Moonlake is building AI for creating world simulations.
ScoutJobs Agent
Get matches like this delivered daily
Sign up free — we'll pull jobs that fit your CV from across the web and rank them for you.
Get started — it's freeMember of Technical Staff - Efficient ML
Moonlake · San Francisco