ML Platform Engineer (Remote — Worldwide)

Join Crusoe’s Infrastructure team to architect the high-performance computing (HPC) software layer that powers large-scale generative AI training on climate-aligned infrastructure. We are building a purpose-built cloud for the LLM era, leveraging stranded energy to fuel massive GPU clusters. As an ML Platform Engineer, you will bridge the gap between bare-metal acceleration and researcher productivity, ensuring seamless orchestration of multi-thousand GPU jobs.

- Architect and maintain the control plane for massive-scale NVIDIA H100 and B200 clusters across carbon-neutral data centers. - Optimize the job scheduling layer to support massive model training runs involving thousands of interconnected GPUs. - Develop custom Kubernetes operators and telemetry pipelines to monitor GPU health, thermal throttling, and interconnect performance. - Collaborate with AI research customers to debug complex distributed training failures and NCCL bottlenecks. - Build and maintain automated provisioning workflows for high-performance storage solutions like Lustre or Weka. - Implement robust fault-tolerance and checkpointing mechanisms to minimize downtime during multi-week training jobs.

- 5+ years of experience in infrastructure engineering, with a focus on ML platforms or distributed systems. - Proven expertise in managing large-scale Kubernetes clusters (EKS, GKE, or bare-metal) and Slurm orchestration. - Deep understanding of high-performance networking fabrics, specifically RoCE v2 or InfiniBand, for low-latency collective communications. - Professional experience with GPU virtualization and container toolkits (NVIDIA Container Toolkit, Enroot, Pyxis). - Proficiency in Go or Python for building platform automation and custom operators. - Familiarity with deep learning frameworks like PyTorch and distributed training libraries such as DeepSpeed or Megatron-LM. - Experience operating in remote-first environments with high degrees of autonomy.

Equity, healthcare, remote stipend.