Solution Architect (AI Infrastructure)

Solution Architect Cluster Design

Company

Sesterce

Conditions

6 days agoLead Full Time Engineering Jobs by Sesterce

Sesterce Stealth Distributed View jobs by Sesterce

Skills

All-Reduce Bandwidth Benchmarking Bom Estimation Cluster Topology Fine-Tuning Flops Utilization Gpu Gpu Cluster Gpu Cluster Design Hpc Inference Infiniband Kubernetes Llm Pre-Training Mfu Ml Model Parallelism Nvidia Gpu Architecture Nvlink Nvswitch Performance Benchmarking Pipeline Parallelism Slurm Technical Pre-Sales Tensor Parallelism Vendor Negotiation Written Communication

About the Role

You will act as the technical bridge between Sesterce's infrastructure capabilities and the compute requirements of enterprise and hyperscale customers, translating AI workloads into precise cluster designs that infrastructure, network, and ops teams can execute against. You will engage directly with customers, produce detailed design documentation, and support commercial negotiations with technical rigor.

Requirements

• 5+ years in HPC or AI infrastructure, with direct involvement in GPU cluster design or technical pre-sales
• Hands-on familiarity with NVIDIA GPU architectures (Hopper, Blackwell), NVLink, NVSwitch, and multi-rail InfiniBand
• Ability to interpret AI workload performance profiles (model parallelism, pipeline vs. tensor parallelism) and map them to hardware
• Strong written communication — design docs must be unambiguous and executable by engineering teams
• Background at or with cloud providers, HPC centers, or AI infrastructure vendors is a strong plus

Responsibilities

• Engage with customers to understand workload profiles (LLM pre-training, fine-tuning, inference) and translate them into cluster sizing and topology recommendations
• Produce cluster design documents covering GPU type (H200, B200, GB300, Blackwell NVL), node count, InfiniBand topology, storage, and interconnect
• Define and validate performance benchmarks (FLOPS utilization, MFU, all-reduce bandwidth) before customer commitment
• Collaborate with the Sesterce OS team to align control plane (Slurm + Kubernetes) with designed topology and scheduling requirements
• Support commercial negotiations with BoM estimates, lead times, and technical risk assessments

Показать контакты