Senior Solutions Architect - AI Infrastructure & Edge Computing - ISV Partners , WWPS Global ISV Partners

**DESCRIPTION** Amazon Web Services (AWS) is seeking a Senior Solutions Architect to join the Worldwide Public Sector ISV Partner Solutions Architecture team, focused on accelerated computing, edge AI, and the open weight model ecosystem. In this role, you will serve as the technical thought leader for strategic partners building GPU-accelerated AI infrastructure and deploying open foundation models at the cloud core, the tactical edge, and in sovereign/air-gapped environments where hosted APIs are not an option.

You will work directly with NVIDIA and adjacent ISV partners to architect full-stack AI compute solutions on AWS, from multi-node GPU training clusters to lightweight inference at the disconnected edge. You will guide partners through model optimization, deployment, orchestration, and infrastructure design patterns that enable mission-critical AI in defense, intelligence community, and national security contexts. This includes scoping, sequencing, and architecting solutions that create measurable mission value while driving clarity across internal AWS teams and external agency stakeholders.

This position requires that the candidate selected be a US Citizen and obtain and maintain an active TS/SCI security clearance.

Key job responsibilities

You Will Own The Technical Relationship With Accelerated Computing And Edge AI Partners Adopting AWS, From First Whiteboard Session To Production At Scale. Specifically, You Will:

• Design and architect GPU-accelerated AI infrastructure on AWS, including multi-node training clusters with EFA networking, high-performance storage, and container orchestration for large-scale model training and inference.
• Guide partners in deploying open weight foundation models (Llama, Mistral, Falcon, Nemotron, Gemma to list a few) across AWS environments, leveraging NVIDIA NIM microservices, Triton Inference Server, TensorRT-LLM, and vLLM for optimized inference at cost and latency targets.
• Architect edge and disconnected compute solutions for DDIL (denied, degraded, intermittent, limited) environments, enabling AI inference on NVIDIA Jetson, IGX, and embedded GPU platforms integrated with AWS hybrid services (Outposts, ECS Anywhere).
• Lead sovereign AI and air-gapped deployment architecture for IL4/IL5/IL6/SCIF environments where partners must run open models entirely on-premise or in isolated AWS regions without reliance on external API endpoints.
• Drive model optimization and right-sizing engagements, including quantization, distillation, pruning, and adapter-based fine-tuning (LoRA/QLoRA) to help partners achieve production-grade performance within infrastructure constraints.
• Serve as the embedded technical advisor for partner engineering teams, conducting architecture reviews, Well-Architected assessments, and proof-of-concept builds that accelerate partner product roadmap delivery on AWS.
• Build executive and working-level relationships with partner CTOs, VP Engineering, and mission-focused government stakeholders (DoD, IC, federal civilian), translating infrastructure capabilities into mission value.
• Develop and publish reusable technical content: reference architectures, blog posts, deployment guides, and benchmark reports for GPU workloads, edge inference patterns, and open model deployment on AWS.
• Identify and drive AWS product feature requests (EC2, Bedrock Custom Model Import, Amazon SageMaker AI, AWS Neuron and AWS Nitro Enclaves) based on partner field signal to influence the service roadmap.
• Establish repeatable benchmarks to evaluate model accuracy, latency, throughput, GPU utilization, reliability, and cost across cloud and edge environments, ensuring solutions meet mission and production requirements.
• Travel approximately 30% of the time for partner site visits, customer engagements, and industry events.

A day in the life

Your morning starts with a design session alongside a partner engineering team, whiteboarding a multi-node GPU training architecture on P5 instances with EFA networking and FSx for Lustre, sizing the cluster for a Llama 3 70B fine-tuning workload that must complete within their sprint cycle.

Mid-morning, you join a call with the partner’s edge deployment team to walk through an inference architecture for NVIDIA Jetson AGX Orin devices operating in a disconnected forward-deployed environment. You map out the model optimization pipeline: quantization via TensorRT-LLM, container packaging, …