Deskripsi Pekerjaan
Nexus Compute is seeking an experienced Staff AI Infrastructure Engineer to design, scale, and optimize bare-metal and cloud GPU clusters running large-scale foundation model workloads. In this role, you will eliminate bottlenecks across compute, InfiniBand networking, and high-performance storage subsystems to accelerate training throughput and lower latency for global inference endpoints.
You will work cross-functionally with machine learning researchers, systems software engineers, and platform teams to build resilient distributed systems that sustain 99.99% reliability at scale.
Tanggung Jawab
- Design, deploy, and scale distributed GPU compute clusters running Kubernetes, Slurm, and Ray.
- Optimize network throughput and latency across high-bandwidth InfiniBand and RoCE fabrics.
- Build automated node-remediation and health-check workflows to minimize cluster downtime during large training runs.
- Collaborate with research teams to profile and eliminate distributed memory and I/O bottlenecks.
- Implement comprehensive observability metrics for GPU utilization, network saturation, and power consumption.
- Drive capacity planning, architectural reviews, and infrastructure cost-reduction initiatives.
Kualifikasi
- 8+ years of systems or infrastructure engineering experience supporting high-concurrency production platforms.
- 3+ years directly managing production GPU infrastructure (NVIDIA H100/A100 clusters).
- Deep expertise in Linux systems internals, kernel tuning, and distributed storage systems (Ceph, Lustre, or WekaIO).
- Strong production engineering capability in Go or Python for tooling, controllers, and automation.
- Demonstrated experience configuring and troubleshooting high-speed networks (InfiniBand, RDMA, RoCEv2).
- Bachelor's degree in Computer Science, Computer Engineering, or equivalent practical experience.