The Role
This is not a generalist SRE role.
You will design, operate, and debug large-scale GPU infrastructure used for distributed training and inference, working directly with customers pushing the limits of modern AI systems.
We’re looking for engineers who have personally run GPU clusters in production, understand the failure modes of distributed training, and can reason about performance from network fabric → kernel → framework.
What You’ll Own
GPU Cluster Architecture: Design and evolve multi-provider, multi-region GPU compute clusters optimized for large-scale training. Make topology-aware scheduling, networking, and storage decisions that directly impact training throughput and cost efficiency.
Customer Technical Partnership: Serve as the primary technical point of contact for customers running large-scale training workloads. Onboard, troubleshoot, and optimize, often in real time.
Reliability & Performance Engineering: Define SLOs and error budgets that account for the unique failure modes of GPU infrastructure (ECC errors, NVLink degradation, NCCL timeouts). Own capacity planning across heterogeneous GPU fleets optimized for training throughput.
Networking & Fabric Health: Ensure the health and performance of high-speed interconnects (InfiniBand, RoCE, NVLink) that underpin distributed training. Diagnose and resolve fabric-level issues that degrade collective operations.
Observability: Build deep visibility into GPU utilization, memory pressure, interconnect throughput, training job performance, and hardware health. Go well beyond standard infrastructure metrics.
Automation & Tooling: Build production-grade automation for cluster provisioning, GPU health checks, job scheduling, self-healing, and firmware/driver lifecycle management.
Incident Leadership: Lead incident response for complex, multi-layer failures spanning hardware, networking, orchestration, and ML frameworks. Drive blameless postmortems and systemic fixes.
What We’re Looking For
GPU Systems Expertise: Deep, hands-on experience operating large-scale GPU clusters (NVIDIA A100/H100/B200 or equivalent). You understand GPU memory hierarchies, ECC behavior, thermal throttling, and hardware failure modes from direct experience not documentation.
High-Performance Networking: Production experience with InfiniBand, RoCE, or NVLink fabrics in the context of distributed training. You can diagnose why an all-reduce is slow, identify a degraded link in a fat-tree topology, and reason about congestion control at scale.
Distributed Training & ML Frameworks: Working knowledge of how large training jobs actually run — NCCL, CUDA, PyTorch distributed, DeepSpeed, Megatron, FSDP, or similar. You don't need to write the models, but you need to understand what's happening at the systems level when a 1,000-GPU training run stalls.
Linux & Systems Internals: Expert-level Linux knowledge: kernel tuning, driver management (NVIDIA drivers, CUDA toolkit), cgroup/namespace internals, performance profiling at the syscall and hardware level.
Kubernetes & Orchestration: Strong experience running Kubernetes in production with GPU workloads, including device plugins, topology-aware scheduling, multi-cluster federation, and custom operators. Experience with Slurm or other HPC schedulers is equally valued.
Automation & Software Engineering: Strong engineering skills in Python, Go, or Bash. You build production-grade tools and services, not just scripts. Infrastructure-as-Code proficiency (Terraform, Helm, Ansible, or equivalent).
Observability & Monitoring: Hands-on experience building monitoring and alerting for GPU infrastructure, not just Prometheus/Grafana basics, but GPU-specific telemetry (DCGM, nvidia-smi, fabric manager metrics) integrated into actionable dashboards.
Incident Management: Proven track record leading incident response for complex distributed systems where the failure could be in hardware, firmware, networking, drivers, orchestration, or application code and you need to narrow it down fast.
Strong Candidates May Have
Distributed Storage: Experience with high-performance parallel file systems (VAST, Weka, Lustre, GPFS) and the checkpoint I/O and data-loading bottlenecks that come with large training runs.
Training Optimization: Experience profiling and optimizing distributed training performance: identifying stragglers, tuning collective communication strategies, improving MFU (Model FLOPs Utilization), and reducing idle GPU time across large runs.
Cluster Buildout & Hardware: Experience involved in physical cluster design - rack layout, power/cooling constraints, network topology design, and hardware validation/burn-in at scale.
Team Leadership: Experience leading or mentoring a team of infrastructure engineers. We're growing and need people who raise the bar for everyone around them.
Why You’ll Love It Here
This is a high-impact, senior builder’s role. You’ll have significant ownership and autonomy to shape how our systems run at a foundational level, working directly with customers and providers while architecting the infrastructure backbone for reliable, scalable AI compute. You’ll influence technical direction and help define what world-class AI infrastructure operations look like.