Overview:
In this role you will design and build the AI data layer for a multi-cloud GPU fleet, enabling high-throughput data paths and scalable storage for thousands of GPUs. You’ll shape Kubernetes-native APIs and storage integrations to make data a first-class citizen in the compute platform. The work supports researchers training NVIDIA’s models, delivering faster training steps and more reliable restarts. You join a mission-driven, builder-focused team that scales AI workloads across clouds and partners. Compensation / Benefits equity benefits
Responsibilities:
Design and build the AI data layer for multi-cloud GPU workloads, including data placement, caching, and data movement across storage backends Create high-throughput data paths for training and reinforcement learning, including streaming loaders and distributed caches Develop Kubernetes-native APIs, operators, and storage integrations for job-declared data on Slurm and Kubernetes Own quotas, capacity, and data lifecycle across hundreds of research teams with attribution, enforcement, retention, and forecasting Implement I/O telemetry and analytics to optimize data placement and diagnose I/O stalls Ensure reliability and incident readiness, with on-call support and preventative changes to reduce data incidents
Key requirements:
BS or MS in relevant field or equivalent experience 5+ years building production systems or data-infrastructure software Experience with parallel or distributed file systems (Lustre, WEKA, VAST, GPFS, Ceph or similar) or large-scale data pipelines and object storage (S3 or compatible) Strong understanding of I/O performance (throughput, latency, caching, metadata) Proficiency in Go, Rust, C or C++, plus Python for tooling Proven track record of owning a service in production and improving it Clear written and spoken communication with engineers and researchers Strong communication Collaborative mindset Analytical thinking Go, Rust, C, or C++; Python Kubernetes and Kubernetes-native APIs; operators Storage systems: Lustre, WEKA, VAST, GPFS, Ceph, S3-compatible object storage