Senior Staff, GPU & Distributed Systems Architect

Reactor; San Francisco (CA); Vor Ort; USD 160.000 - 230.000; Vollzeit

Zusammenfassung

Reactor is building the inference platform for real-time world models in San Francisco. The team owns multi-node inference, fast startup, and GPU-accelerated serving across Kubernetes clusters and cloud providers.

You will work on model distribution, scheduling, and GPU communication to cut cold starts, reduce latency, and improve throughput. Strong systems fundamentals and experience with GPUs are essential for this hands-on role.

Qualifikationen

• 3+ years building distributed systems, cloud platforms, or performance-sensitive systems software in production.
• Proficient in Go, Rust, C++, or Python with a solid understanding of concurrency and resource management.
• Hands-on GPU or ML systems experience, such as model loading, accelerator memory management, distributed execution, or inference serving.
• Strong Linux, networking, and storage fundamentals, with experience diagnosing performance across software and hardware boundaries.
• Experience extending Kubernetes through controllers, operators, or scheduling plugins, or building comparable orchestration systems.

Aufgaben

• Multi-node inference: build distributed worker initialization, sharded model loading, and coordinated execution across GPUs and machines.
• Fast model startup: engineer parallel checkpoint loading, weight streaming, host-to-device transfers, and GPU snapshot/restore to cut cold starts.
• Model distribution: build transfer, caching, prefetching, and replication systems that move weights and large container images across clusters, regions, and clouds.
• Scheduling and orchestration: develop Kubernetes controllers, scheduling extensions, and placement algorithms for model residency and latency across clouds.
• GPU communication: optimize NCCL/NVLink usage and RDMA to reduce synchronization and overlap with computation.
• Reliable, efficient serving: build admission control, backpressure, and recovery from worker failures across accelerators.
• Benchmarks: create reproducible tests and measure startup time, latency, throughput, and fleet utilization.

Kenntnisse

Go

Rust

C++

Python

Distributed systems

Tools

Kubernetes

NCCL

RDMA

CUDA

TensorRT

PyTorch

Jobbeschreibung

Reactor is building the inference platform for real-time world models in San Francisco. The team owns multi-node inference, fast startup, and GPU-accelerated serving across Kubernetes clusters and cloud providers.

You will work on model distribution, scheduling, and GPU communication to cut cold starts, reduce latency, and improve throughput. Strong systems fundamentals and experience with GPUs are essential for this hands-on role.