Deskripsi Pekerjaan
NeuralScale Systems is seeking a visionary Staff AI Infrastructure Engineer to architect, scale, and optimize the distributed training and inference platforms powering our foundation models. In this role, you will lead the core systems team responsible for managing heterogeneous GPU clusters, cutting-edge networking fabrics, and low-latency storage engines. The ideal candidate blends deep distributed systems fundamentals with hands-on experience scaling large-scale ML infrastructure.
Tanggung Jawab
- Architect, deploy, and scale high-throughput GPU clusters supporting multi-node, distributed model training workflows.
- Build resilient, fault-tolerant orchestration layers using Kubernetes and Slurm to maximize compute utilization.
- Optimize network architectures, leveraging RoCEv2, InfiniBand, and NCCL primitives for distributed communications.
- Design automated health checking, fault detection, and recovery pipelines to handle silent data corruption and hardware degradation.
- Collaborate with ML research teams to eliminate training pipeline bottlenecks across storage, I/O, and inter-node bandwidth.
- Champion site reliability engineering (SRE) practices, setting observability standards and SLAs for core model deployment services.
Kualifikasi
- Bachelor’s or Master’s degree in Computer Science, Distributed Systems, or equivalent practical experience.
- 8+ years of core software engineering experience, with at least 4 years focused on large-scale distributed systems or ML platforms.
- Demonstrated proficiency in Go, Python, C++, or Rust.
- Deep hands-on experience with container orchestration (Kubernetes internals), GPU scheduling, and Linux kernel performance tuning.
- Direct experience profiling and optimizing distributed ML frameworks such as PyTorch, DeepSpeed, or Megatron-LM.
- Strong background with high-performance storage systems (Ceph, Lustre, GPFS) and high-speed networking (InfiniBand, RDMA).