Staff Foundation Model Inference Engineer

United States Digital Space LLC; San Francisco (CA); On-site; USD 190,000 - 265,000; Full time

Job summary

United States Digital Space LLC is seeking seasoned backend/infrastructure engineers to build and optimize LLM inference platforms powering enterprise-grade AI workloads. You will tackle reliability, latency, and scalability challenges across distributed systems and cloud-native infrastructure.

Join a high-visibility Foundation Model Inference team building enterprise-ready APIs with governance and scalable deployment for model serving, training, and vector search.

Qualifications

• 8+ years of experience in backend or infrastructure engineering.
• Experience with distributed systems, scalable APIs, or cloud-native infrastructure.
• Experience with real-time serving, ML infrastructure, or GPU orchestration.
• Familiarity with service-oriented architecture, deployment pipelines, and system observability.

Responsibilities

• Build LLM infrastructure powering large-scale inference workloads for customers through partner models (OpenAI, Anthropic, Gemini) and self-hosted models (Qwen, GPT-OSS, Llama).
• Improve reliability, latency, and efficiency of distributed AI workloads.
• Collaborate with platform, infra, and ML teams to deliver seamless end-to-end experiences.
• Shape how developers and data scientists build and interact with AI on the company.

Skills

Distributed systems

Cloud-native infrastructure

Real-time serving

Observability

Tools

SageMaker

Vertex AI

Azure ML

MLflow

PyTorch

Ray

vLLM

SGLang

Job description

United States Digital Space LLC is seeking seasoned backend/infrastructure engineers to build and optimize LLM inference platforms powering enterprise-grade AI workloads. You will tackle reliability, latency, and scalability challenges across distributed systems and cloud-native infrastructure.

Join a high-visibility Foundation Model Inference team building enterprise-ready APIs with governance and scalable deployment for model serving, training, and vector search.