This is a fully remote position, open to applicants in California.

📋 Description

• Collaborate with inference partners to develop state-of-the-art inference services, demonstrating the advantages of NVIDIA's technology and gathering product feedback.

• Create and manage inference recipes utilizing NVIDIA Dynamo, allocating tasks among GPU workers to enhance operational efficiency.

• Speed up inference pipelines with TensorRT-LLM, vLLM, SGLang, and other backend technologies.

• Advocate for DevOps best practices in the management of Kubernetes clusters, configuration of computing fabrics, and ensuring observability.

• Offer mentorship and technical guidance to both customers and internal teams.

• Oversee the deployment of disaggregated inference systems and troubleshoot complex challenges.

• Draft reference architectures implemented by over 10,000 GPU clusters.

• Instruct C-level executives, distinguished engineers, and datacenter developers on scaling businesses, systems, and infrastructure.

⛳️ Requirements

• Over 6 years of experience in Solutions Architecture (or related fields like Solution Engineering or Forward-Deployed Engineering) driving customer interactions for deploying distributed systems, including at least 2 years focused on AI workloads within Kubernetes.

• Proficiency with at least one of the following: NVIDIA Dynamo, Triton Inference Server, or TensorRT-LLM for model optimization and deployment.

• Extensive understanding of contemporary inference best practices, including disaggregated serving, multi-tier key-value cache management, speculative decoding, quantization, and custom inference kernels.

• Practical experience in full-stack agent design utilizing modern sandboxing, memory retrieval systems, along with evaluation, skill design, and governance.

• Bachelor’s degree in Computer Science/Engineering or equivalent professional experience.

• Prior experience with the deployment of NVIDIA inference technologies such as Dynamo, NIXL, and Grove.

• Familiarity with model customization techniques including SFT, DPO, GRPO, RLVR, and the associated data curation strategies.

• Contributions to relevant open-source projects.