Zipporin is seeking a talented and driven MLOps & Distributed GPU Systems Engineer to join our team in Austin, TX on a hybrid schedule. In this role, you will architect, optimize, and maintain high-performance compute infrastructure that accelerates large-scale model training and production inference. You will work closely with our research and machine learning engineering teams to eliminate training bottlenecks, scale GPU clusters, and streamline our automated deployment pipelines.
Key Responsibilities:
- Design, scale, and manage multi-node distributed GPU clusters leveraging Kubernetes, Ray, and Slurm across hybrid cloud and bare-metal environments.
- Optimize distributed deep learning workflows using frameworks such as PyTorch, DeepSpeed, Megatron-LM, and vLLM.
- Build and automate robust MLOps CI/CD pipelines for continuous integration, model evaluation, artifact tracking, and deployment.
- Profile and debug hardware-level and networking bottlenecks (NCCL, InfiniBand, RoCE, CUDA kernels) to maximize GPU utilization.
- Implement comprehensive observability, telemetry, and fault-detection platforms for large-scale GPU workloads using Prometheus, Grafana, and custom tooling.
Qualifications:
- 5+ years of experience in systems engineering, DevOps, or infrastructure engineering, with at least 3 years dedicated to ML/AI infrastructure.
- Deep understanding of NVIDIA GPU architecture, CUDA runtimes, and high-speed interconnects (InfiniBand, NVLink, NCCL).
- Demonstrated experience scaling distributed training frameworks (FSDP, DeepSpeed, Megatron) across multi-node topologies.
- Proficiency in containerization and orchestration tooling, specifically Kubernetes, Docker, and workload schedulers like Slurm or Ray.
- Strong software engineering background in Python and Go, alongside Infrastructure as Code (Terraform, Ansible).
- Bachelor's or Master's degree in Computer Science, Computer Engineering, or equivalent practical experience.