AI Accelerator System Architect

About this role

The era of pervasive AI has arrived. In this era, organizations will use generative AI to unlock hidden value in their data, accelerate processes, reduce costs, drive efficiency and innovation to fundamentally transform their businesses and operations at scale.

SambaNova Suite™ is the first full-stack, generative AI platform, from chip to model, optimized for enterprise and government organizations. Powered by the intelligent SN40L chip, the SambaNova Suite is a fully integrated platform, delivered on-premises or in the cloud, combined with state-of-the-art open-source models that can be easily and securely fine-tuned using customer data for greater accuracy. Once adapted with customer data, customers retain model ownership in perpetuity, so they can turn generative AI into one of their most valuable assets.

About the team

The System Architecture team owns architecture for RDU-based platforms at the board, rack, and cluster level, covering SambaRack-class server and rack design, the scale-up and scale-out interconnect fabric, power delivery, and air and liquid cooling. We write the specifications ODM/OEM and component partners build to, evaluate their proposals, and make the build, buy, and co-develop calls. We define the bring-up, validation, and qualification plan for each platform, and work with signal and power integrity engineers, SambaFlow, ML performance, and datacenter operations to keep architecture decisions matched to the workloads running on the systems.

About the role

SambaNova's Reconfigurable Dataflow Unit (RDU) powers a full-stack, chip-to-model AI platform used by enterprise and government customers, delivered both on-premises and in the cloud. As a Senior RDU System Architect, you'll be a part of the team that turns silicon into shippable systems — SambaRack-class server and rack designs, the scale-up/scale-out fabric that links RDUs together, and the power and cooling envelope that lets dense AI compute run efficiently in a real data hall. You'll set direction across build, buy, and co-design decisions with our hardware partners, write the specifications those partners build to, and stay engaged from architecture definition through bring-up, qualification, and fleet deployment.

Responsibilities

Some of your responsibilities will include:

Platform & Rack Architecture

• Being a part of the team defining the architecture of RDU-based server and rack platforms, from board layout and chassis design through full-rack integration, translating product and customer requirements into system specifications that engineering teams and vendors can build to.

• Select and validate major system components — RDU sockets, memory modules, power supplies, chassis mechanicals — balancing performance, power envelope, and manufacturability.

• Build performance and cost models to compare candidate rack configurations before committing to a design.

Fabric, Power & Thermal

• Architect the scale-up and scale-out interconnect fabric that links RDUs into larger training and inference clusters, choosing topology, switch/NIC hardware, and optics to hit bandwidth and latency targets.

• Own the power delivery architecture (including high-density/48V-class distribution) and the cooling strategy — air and liquid — needed to keep dense, multi-socket racks within thermal budget.

• Work with signal and power integrity engineers to confirm interconnect and power designs hold up once they leave simulation and reach the lab.

Partner & Supply Chain Engineering

• Evaluate ODM/OEM and component vendor proposals against SambaNova's system requirements, and decide where to build in-house, buy off the shelf, or co-develop with a partner.

• Flag schedule or technical risk in partner designs early, and drive the changes needed to close the gap.

• Maintain a working map of the vendor and technology landscape — boards, interconnect, power, cooling — that feeds SambaNova's platform roadmap.

Bring-Up, Qualification & Field Reliability

• Define the bring-up, validation, and qualification plan for each new platform, including the reliability (RAS) and telemetry targets it needs to hit once it's running in a customer's data hall.

• Lead root-cause investigations when a platform's measured behavior in the lab or field diverges from the architecture's projections.

Cross-Team Technical Leadership

• Partner with the SambaFlow software/compiler team, ML performance, and datacenter operations so system architecture choices actually help the workloads running on top of them.

• Brief executive leadership on architecture tradeoffs and keep the roadmap aligned to customer and product priorities.

• Mentor engineers on system architecture methodology, and help build the review process and spec standards the team uses going forward.

Required Qualifications

• B.S. or M.S. in Electrical Engineering, Computer Engineering, Computer Science, or equivalent practical experience

• 12+ years architecting hardware systems for hyperscale, HPC, or AI/ML infrastructure

• Deep, hands-on background in at least one systems domain, such as interconnect, power delivery, thermal/cooling, or mechanical, for large-scale AI accelerator, HPC, or hyperscale systems

• Experience owning system architecture at board, rack, or cluster scale, including writing the specs that other engineering teams and vendors build to

• A track record of carrying a system through architecture definition, bring-up, and volume production

• Direct experience holding external vendors and ODM/OEM partners accountable to a technical specification

• Comfortable reasoning across domains that affect each other, for example how an interconnect choice changes the power and thermal budget, which in turn changes the mechanical design

• A history of making and owning directional calls, and building alignment across engineering and partner teams to support them

• Clear technical writing, specs and reviews other teams will build to and rely on

Preferred Qualifi

Similar jobs

• AI Application Architect — Procore · Austin, Texas, United States
• Accelerator Architect — Palo Alto Networks · Remote
• ML Accelerator Architect — Waymo · Mountain View, CA, US; New York City, NY, US
• AI Growth Architect — Hilberts · San Francisco
• AI Growth Architect — Newsbreak · Mountain View, California, United States
• AI Architect — Accenturefederalservices · Chantilly, VA
• AI Architect — "Stride · Remote
• AI Architect — Change · Remote