LLM Inference & GPU Systems Consultant
Boost your chances before you apply.
Delan Associates, Inc
Charlotte, NC, US
Summary
The consultant will manage and optimize large-scale on-prem LLM inference infrastructure using NVIDIA H200 clusters and OpenShift AI. Responsibilities include driving runtime efficiency, managing inference engines like vLLM, and overseeing the complete Hugging Face model lifecycle.
Job Description
Job Title: LLM Inference & GPU Systems Consultant
Location: Charlotte, NC (Onsite)
Duration: 6+ Months
Must be onsite at client in Charlotte, NC at least 3 days/week
Role Overview:
We are seeking an AI Infrastructure Runtime Engineer to build and maintain large-scale on-prem LLM infrastructure. This is an enterprise private GenAI environment running on NVIDIA H200 GPU clusters and an OpenShift AI deployment ecosystem. You will manage production inference internally, including self-hosting open-source LLMs like Llama. We are focused exclusively on inferencing; this role involves no model training infrastructure or fine-tuning pipelines.
Key Responsibilities
NVIDIA GPU Runtime Optimization: Drive extreme runtime efficiency and optimization for the token generation pipeline. Specifically manage prefill/decode optimization and KV cache management.
Inference Serving: Deploy and manage inference engines including vLLM and TensorRT-LLM.
Hardware Utilization: Optimize GPU throughput tuning, batching strategies, and latency optimization. Manage workload orchestration using RunAI and Kubernetes GPU orchestration.
Model Lifecycle Management: Oversee the complete Hugging Face model lifecycle, including model onboarding, deployment, and retirement.
Platform Operations: Operate and maintain the OpenShift AI ecosystem as the primary container platform for GenAI workloads.
Required Qualifications
8+ years experience working as an LLM Systems Engineer or AI Infrastructure Runtime Engineer.
8+ years hands-on experience with NVIDIA H200 clusters and runtime optimization techniques (KV Cache, prefill/decode).
Proficiency in OpenShift AI and GPU orchestration tools like RunAI.
Strong experience with modern inference frameworks, specifically vLLM and TensorRT-LLM.
Proven track record managing the Hugging Face deployment lifecycle.
