Principal Solutions Architect
$170k - $190k/yr+ Commission + Discretionary bonus
Expert
Bachelor's, Master's
San Ramon, CA, USA
In Person
Solution EngineeringPre-Sales Solution Engineering
About the job
Requirements
- Bachelor's degree in Computer Science, Electrical Engineering, Computer Engineering, or a related technical discipline; a master's degree is preferred.
- At least 8 years of solutions architecture, systems engineering, or technical pre-sales experience, including at least 4 years focused on GPU infrastructure or high-performance computing environments.
- A proven track record designing and deploying NVIDIA DGX or HGX-based GPU clusters in production artificial intelligence or machine learning environments.
- Deep understanding of distributed deep learning concepts, including tensor parallelism, pipeline parallelism, data parallelism, gradient checkpointing, and mixed-precision training.
- Hands-on experience with InfiniBand or high-speed Ethernet fabric design, RDMA configuration, and collective communication tuning using NCCL or MPI.
- Direct experience sizing and deploying parallel storage systems such as VAST, Hammerspace, or Lustre, WEKA, or GPFS for AI training workloads.
- Working knowledge of Kubernetes, GPU Operator, and at least one GPU workload scheduler such as Run:ai or SLURM.
- Experience with Linux system administration, CUDA development environment configuration, and GPU driver and firmware management.
- Ability to create technical proposals, architecture diagrams using Visio, Lucidchart, or draw.io, and bill-of-materials-level documentation.
- Ability to present to technical audiences and C-level executives.
Responsibilities
- Lead discovery workshops to capture AI and machine learning workload requirements, including model training scale, inference service-level agreements, data pipeline throughput, and multi-tenancy needs.
- Architect full-stack AI Factory solutions aligned to NVIDIA reference architectures, integrating colocation, GPU compute, networking, storage, and software layers.
- Develop bills of materials, rack elevation diagrams, network topology drawings, and power and cooling budgets for customer proposals.
- Define GPU cluster architectures using NVIDIA DGX, HGX, and MGX systems with B200, B300, and GB300 Blackwell SXM and NVLink-Switch configurations.
- Design RTX PRO 6000 Blackwell Server Edition deployments for inference-optimized and enterprise AI workloads.
- Conduct workload sizing and total-cost-of-ownership and return-on-investment modeling to validate infrastructure dimensioning for training, fine-tuning, and inference at scale.
- Specify colocation requirements, including critical power load, UPS and generator configurations, and power usage effectiveness targets.
- Design high-density GPU deployments using air cooling, direct liquid cooling, and rear-door heat exchanger configurations.
- Define meet-me room and cross-connect requirements and specify carrier-neutral telecommunications diversity strategies.
- Engage colocation providers and data center operators to validate capacity availability and negotiate technical service-level agreements.
- Coordinate with facilities and mechanical, electrical, and plumbing engineers to validate power infrastructure from utility feed through power distribution unit to rack level.
- Architect multi-node GPU clusters optimized for large language model pre-training, fine-tuning, and reinforcement learning from human feedback.
- Size and configure DGX SuperPOD, HGX H- and B-series, and MGX modular systems based on model parameter count, dataset size, and iteration timelines.
- Define server firmware, BIOS, BMC, and DGXOS baselines for production GPU infrastructure.
- Establish GPU health monitoring, reliability, availability, and serviceability policies, and lifecycle management procedures.
- Design backend GPU fabric networks using NVIDIA Quantum InfiniBand NDR 400 Gb/s and HDR 200 Gb/s for distributed training traffic.
- Architect Spectrum-X Ethernet-based AI networking solutions for inference clusters requiring high-bandwidth, low-latency connectivity.
- Specify ConnectX-8 and ConnectX-7 host channel adapter deployments and configure RDMA over Converged Ethernet or InfiniBand transport for NCCL collective operations.
- Integrate BlueField-3 DPUs for GPU-accelerated network functions, storage offload, zero-trust security isolation, and bare-metal provisioning.
- Design leaf-spine and fat-tree topologies for non-blocking bisectional bandwidth in GPU training clusters.
- Define quality-of-service policies separating storage, compute fabric, and management-plane traffic.
- Design high-performance parallel file-system solutions using VAST Data, Hammerspace, and Pure Storage FlashBlade//E for AI training and checkpoint storage.
- Size storage capacity, IOPS, and throughput based on dataset characteristics, checkpoint frequency, and concurrent reader and writer counts.
- Architect multi-tier storage hierarchies with hot NVMe flash for active datasets, warm object storage for model archives, and cold tape or cloud for long-term retention.
- Configure VAST Data Universal Storage for disaggregated storage with NFS, S3, and POSIX access, and tune it for large sequential-read performance.
- Deploy Hammerspace Global Data Environment for distributed data management and NFS-over-RDMA acceleration across geographically dispersed GPU clusters.
- Define data pipeline architectures ingesting from cloud object stores such as S3, GCS, and ABS to local flash for GPU-local data loading without I/O bottlenecks.
- Deploy and configure the NVIDIA AI Enterprise software stack, including NVIDIA GPU Operator, NIM microservices, and RAPIDS accelerated data science libraries.
- Architect inference-serving infrastructure using NVIDIA NIM for optimized large language model and vision model deployment with autoscaling.
- Implement NVIDIA Dynamo for distributed inference and disaggregated serving of large-scale generative AI models.
- Configure and optimize the CUDA toolkit, cuDNN, NCCL communication libraries, and custom kernel environments for training workloads.
- Deploy Base Command Manager and DGXOS for cluster lifecycle management, node provisioning, health dashboards, and job-scheduling integration.
- Integrate NVIDIA Mission Control for AI Factory operations, observability, and multi-cluster fleet management.
- Design and deploy Kubernetes-based AI platforms using NVIDIA GPU Operator, integrating Run:ai for dynamic GPU resource scheduling and multi-tenant workload isolation.
- Configure SLURM workload manager for HPC-style job scheduling on bare-metal GPU clusters, including preemption policies, fair-share scheduling, and burst-to-cloud integration.
- Establish MLOps toolchain integrations with frameworks including PyTorch, JAX, and TensorFlow, and experiment-tracking platforms including MLflow and Weights & Biases.
- Serve as the primary technical point of contact throughout the pre-sales and delivery lifecycle, from initial discovery through post-deployment optimization.
- Produce and present architecture design documents, technical proposals, and executive-level briefings to CTO, CIO, and VP-level stakeholders.
- Lead proof-of-concept and pilot deployments, including benchmark design, execution, and results analysis.
- Collaborate with procurement, logistics, and deployment teams to ensure on-time delivery of complex infrastructure programs.
- Provide post-deployment hypercare support, performance tuning, and capacity-planning advisory services.
- Contribute to internal knowledge bases, solution playbooks, and reference architectures for repeatable AI Factory deployments.
Desired Qualifications
- NVIDIA-certified professional credentials such as DCA-Core, NCP-DS, or equivalent.
- Experience with NVIDIA Base Command Platform or Mission Control for multi-cluster AI Factory operations.
- Familiarity with sovereign AI, government cloud, or regulated-industry AI infrastructure requirements.
- Experience integrating AI Factory infrastructure with public cloud platforms such as AWS, Azure, and GCP for hybrid and burst-to-cloud architectures.
- Background in MLOps, LLMOps, or platform engineering for production AI model lifecycle management.
- Prior experience with colocation data center procurement, RFP development, and SLA negotiation.
- Contributions to open-source AI infrastructure projects or published technical content such as blogs, whitepapers, or conference presentations.
- Active participation in the NVIDIA Partner Network ecosystem or prior experience at an NVIDIA Elite Solution Provider.
About the company
ePlus provides security, cloud, and managed IT services to help organizations secure, modernize, optimize, and scale their IT infrastructure. It offers data migration, hosting and management, data center, collaboration, and networking services, backed by a large team of technical experts and numerous certifications. The company works through a services-led approach that combines strategy, execution, and measurable results to help over 4,300 global customers navigate complex IT, regulatory, and workforce environments. Compared with competitors, ePlus differentiates itself with its extensive expert bench, deep certifications, long-standing industry presence (30+ years), and a proven focus on customer outcomes. Its goal is to deliver transformational IT results by enabling secure, efficient, and scalable technology environments for its clients.
Company Size
1,001-5,000
Company Stage
IPO
Headquarters
Herndon, Virginia
Founded
1990
Get referred to ePlus
See people who can refer or advise you
Simplify's Take
What believers are saying
- Q1 fiscal 2027 net sales rose 1.0% to $649.1 million on August 4, 2026.
- Management cited record sales and significantly higher booked and open orders.
- Everpure and HPE partnerships expand storage, maintenance, and managed-services cross-sell.
What critics are saying
- Memory chip shortages delayed shipments and extended lead times in Q1 2027.
- Fiscal 2026 cash conversion cycle worsened to 51 days from 29.
- Microsoft ecosystem dependence makes channel-rule changes an existential threat.
What makes ePlus unique
- ePlus became North America's first Everpure Ambassador partner on September 2, 2026.
- Daymark adds Microsoft Tier 1 CSP depth and Boston-area enterprise reach.
- Its services-led model spans security, cloud, data center, networking, and managed services.
Help us improve and share your feedback! Did you find this helpful?
Benefits
- Health Insurance
- Employee Stock Purchase Plan
- Paid Vacation
- Paid Sick Leave