Teams at Sciforium

Design, implement, and optimize custom GPU kernels for large-scale AI training and inference. Integrate low-level kernels with PyTorch, JAX, and internal runtimes; profile workloads, identify bottlenecks, develop performance models, and improve accelerator efficiency. Collaborate with ML researchers, distributed systems engineers, model-serving teams, and NVIDIA/AMD hardware partners while contributing to benchmarking, testing, tooling, and documentation.

Owns the company’s on-site office operations, including facilities, supplies, inventory, vendors, contractors, AV support, access badges, equipment, new-hire setup, events, executive logistics, and office expansion. The role also manages office budgets, develops operational playbooks, and ensures a welcoming, functional workplace. Requires strong organization, communication, vendor and budget management, basic IT skills, and hands-on problem-solving. This is an on-site San Francisco role with occasional early or late hours.

Owns the reliability, performance, and operation of a high-density GPU compute cluster. Responsibilities include Linux administration, cluster monitoring, GPU deployment and driver management, kernel and hardware debugging, network security, identity and distributed storage administration, and maintenance of CUDA, ROCm, PyTorch, JAX, vLLM, and related ML infrastructure. The role also handles incident response, vendor coordination, performance tuning, and automation for large-scale AI infrastructure.