Deskripsi Pekerjaan
Nexus Scale AI is seeking an exceptional Senior AI Infrastructure Engineer to lead the design, scaling, and operationalization of our high-throughput distributed training and inference platforms. In this role, you will build the computational backbones that power cutting-edge foundation models deployed directly to enterprise clients worldwide.
We provide a fast-moving, engineering-first environment with access to massive GPU clusters, autonomous decision-making, and high-impact initiatives. You will work closely with research scientists and platform architects to eliminate bottlenecks, minimize latencies, and optimize resource efficiency across multi-cloud environments.
Tanggung Jawab
- Architect, deploy, and maintain large-scale Kubernetes clusters optimized for GPU-accelerated distributed deep learning workloads.
- Design fault-tolerant training pipelines utilizing PyTorch, Ray, and high-performance interconnects (InfiniBand/RoCE).
- Build automated observability, monitoring, and telemetry platforms for hardware health and utilization metrics using Prometheus and Grafana.
- Implement ultra-low-latency model inference runtimes leveraging Triton Inference Server, TensorRT, and vLLM.
- Automate infrastructure provisioning via Infrastructure-as-Code (Terraform) across AWS and bare-metal environments.
- Collaborate with research teams to profile and optimize CUDA kernels, memory bandwidth, and distributed checkpointing mechanisms.
Kualifikasi
- 5+ years of production experience in infrastructure engineering, distributed systems, or MLOps roles.
- Proficiency in Go, Python, and shell scripting within production Linux operating environments.
- Demonstrated experience running large-scale workloads on Kubernetes using custom operators and CRDs.
- Hands-on experience managing NVIDIA GPU infrastructure, CUDA drivers, and networking fabrics at scale.
- Deep understanding of distributed training frameworks such as PyTorch FSDP, DeepSpeed, or Megatron-LM.
- B.S. or M.S. in Computer Science, Computer Engineering, or equivalent practical experience.