- Bellevue, Washington, United States
- Hybrid
- Full-time
Build and operate production-grade model-serving and inference systems supporting high-throughput, low-latency AI workloads.
Compensation
Competitive base pay for Bellevue market Certain roles are eligible for additional rewards, including merit increases, annual bonus, and long term incentives.
Responsibilities
- Build and operate production-grade model-serving and inference systems supporting high-throughput, low-latency AI workloads.
- Optimize inference infrastructure for token throughput, latency, scalability, and cost efficiency across different model architectures and workloads.
- Design systems that maximize GPU utilization while maintaining predictable performance and reliability.
- Improve the scalability and operational maturity of inference platforms as customer demand grows.
- Partner with AI training, GPU performance, orchestration, and infrastructure teams to ensure smooth transitions from model development to production serving.
Requirements
- Nice to have: vLLM
- Nice to have: SGLang
- Nice to have: TensorRT-LLM
- Nice to have: Triton Inference Server
- Nice to have: LLM inference