Full-Time
AI Accelerator System Architect
Posted on 8/18/2026
SambaNova Systems
201-500 employees
Enterprise AI platform with hardware stack
Compensation Overview
$250k - $350k/yr
San Jose, CA, USA
In Person
Bachelor's, Master's
 Hardware Engineering (1) |
|---|
Get referred to SambaNova Systems
See people who can refer or advise you
- A Bachelor of Science or Master of Science degree in Electrical Engineering, Computer Engineering, or Computer Science, or equivalent practical experience.
- 12+ years of experience architecting hardware systems for hyperscale, high-performance computing, or artificial intelligence/machine-learning infrastructure.
- Deep, hands-on experience in at least one systems domain, such as interconnect, power delivery, thermal/cooling, or mechanical systems, for large-scale AI accelerator, HPC, or hyperscale systems.
- Experience owning system architecture at board, rack, or cluster scale, including writing specifications that 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.
- Ability to reason across interdependent domains, including how interconnect choices affect power, thermal, and mechanical design.
- A history of making and owning directional decisions and building alignment across engineering and partner teams.
- Clear technical writing, specifications, and reviews that other teams can build to and rely on.
- Define 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.
- Select and validate major system components, including RDU sockets, memory modules, power supplies, and chassis mechanicals, while balancing performance, power envelope, and manufacturability.
- Build performance and cost models to compare candidate rack configurations before committing to a design.
- Architect the scale-up and scale-out interconnect fabric linking RDUs into larger training and inference clusters, selecting topology, switch/NIC hardware, and optics to meet bandwidth and latency targets.
- Own the power delivery architecture, including high-density/48V-class distribution, and the air and liquid cooling strategy for dense, multi-socket racks.
- Work with signal and power integrity engineers to validate interconnect and power designs in the laboratory.
- Evaluate ODM/OEM and component-vendor proposals against system requirements and decide what to build in-house, buy off the shelf, or co-develop with partners.
- Identify schedule or technical risks in partner designs and drive changes needed to close gaps.
- Maintain a working map of the vendor and technology landscape covering boards, interconnect, power, and cooling to inform the platform roadmap.
- Define bring-up, validation, and qualification plans for each new platform, including reliability and telemetry targets for customer data halls.
- Lead root-cause investigations when measured platform behavior in the laboratory or field diverges from architectural projections.
- Partner with the SambaFlow software/compiler team, ML performance, and datacenter operations to align system architecture with workloads.
- Brief executive leadership on architecture tradeoffs and keep the roadmap aligned with customer and product priorities.
- Mentor engineers on system architecture methodology and help establish the team’s review process and specification standards.
- Hands-on experience with AI accelerator platforms such as RDU, GPU, TPU, or custom ASIC systems and the fabrics that scale them up and out.
- Fluency in chip- and package-level design sufficient to work as a peer with silicon and packaging architects.
- Experience establishing a new engineering function, review process, or design practice from scratch.
Generating company summary.
Company Size
201-500
Company Stage
Series F
Total Funding
$2.5B
Headquarters
Palo Alto, California
Founded
2017
Get referred to SambaNova Systems
See people who can refer or advise you
Simplify's Take
What believers are saying
- July 2026 Series F raised $1 billion at $11 billion, extending runway.
- JPMorganChase selected SambaNova in July 2026 for secure on-prem enterprise inference.
- August 4, 2026 added Mohsen Moazami to expand sovereign and government sales.
What critics are saying
- April 2025 layoffs cut 77 employees after SambaNova abandoned training-cloud ambitions.
- Nvidia, Groq, and custom ASIC rivals compress pricing and customer patience through 2027.
- If SN50 shipments slip, the $3.5 billion VC2 order and 2027 IPO thesis weaken.
What makes SambaNova Systems unique
- SambaNova sells inference-optimized RDUs and SN50 systems, not GPU-first training clusters.
- On-prem deployments keep sensitive banking and sovereign data inside customer infrastructure.
- Intel collaboration and heterogeneous Nvidia-plus-SambaNova architectures widen deployment flexibility.
Help us improve and share your feedback! Did you find this helpful?
Benefits
Flexible PTO in US
Parental Leave
Benefits (medical, dental and vision)
Flexible Spending Accounts
401k/Pensions
Gym Access
Flexible Working Hours
Growth & Insights and Company News
6 month growth
↓ -1%1 year growth
↓ -4%2 year growth
↑ 0%Callosum raises $100M led by Atomico in one of Europe's largest seed rounds. August 20, 2026 * Having raised $10.25 million in a pre-seed round in February, Callosum has now obtained $100 million in seed funding. * Atomico led the seed round, with Plural, DCVC, and the UK's Sovereign AI Fund also joining. * According to the fund's website, the investment in Callosum was its first, made in April. Callosum, a London-based startup, has raised a $100 million seed round, one of the largest ever in Europe, led by Atomico. This follows its pre-seed round at a $10.25 million valuation led by Plural in February. UK's £500 million Sovereign AI Fund, along with some other unnamed investors and angels, took part. Since its foundation in 2025, Callosum has now raised approximately $110 million in funding. The company has not shared its valuation. From neuroscience PhDs to a nine-figure seed. Callosum was founded by Danyal Akarca and Jascha Achterberg, who first began working together while pursuing their PhDs at Cambridge in neuroscience, computing, and AI. Their research has appeared in Nature journals, and both have held positions at Intel and Google DeepMind. Callosum's method is based on the founders' academic backgrounds. They hold the view that, just as intelligence in nature arises from a variety of neurons, AI should not rely on identical chips. Instead, their software breaks down AI tasks into separate steps and sends each one to the most suitable model and chip rather than using the same hardware for all purposes. As a growing share of industry spending shifts from training to inference, AI companies frequently spend half or more of their revenue on inference. For instance, one step in an autonomous computing task may require fast pattern matching, while another may call for deeper reasoning. The company's system therefore allocates each step to different hardware, thus preventing competition for resources. It claims that its solution is twice as accurate, seven times faster, and four times cheaper for complex tasks compared with using uniform hardware, even though it has not made the technical benchmark figures available. Nvidia's CUDA ecosystem, with its ~85% share of the GPU market, remains Callosum's principal competitor. Cerebras, the wafer-scale chipmaker that went public in May 2026 at a valuation of nearly $56 billion, has now entered into a partnership with Callosum. Although Etched and SambaNova are also developing their own custom chips to improve inference efficiency, their methods are less extensive than Callosum's chip-agnostic approach. A 10x increase in funding since February. The company's new funding will enable it to become a "global heterogeneous integrator" and establish new computing partnerships. Its key partnership with Cerebras focuses on achieving low-latency inference at scale, and it is also entering into a collaboration with the Korean chipmaker Rebellions. "By integrating Cerebras into Callosum's platform, we're making ultra-low-latency inference available exactly where it creates the greatest impact, enabling customers to build AI systems that simply weren't practical before," said Andrew Feldman, Cerebras' chief executive. Sunghyun Park, chief executive of Rebellions, framed the deal as a structural stand against single-vendor lock-in: "Working with Callosum puts our architecture into systems alongside hardware chosen for different parts of the workload, instead of asking one chip to do every job. That's the difference between a partnership and a deployment that only works in one environment or geography." Kanishka Narayan, the UK's minister for artificial intelligence, tied the investment to national chip strategy: "AI is nothing without the chips that underpin it, and the eye-watering demand for them is only going to grow. In the race to develop and use AI, success will depend not just on having access to those chips, but on using them as efficiently as possible." Nvidia believes the total AI infrastructure market could reach at least $1 trillion by 2027, driven primarily by inference workloads. Callosum maintains that no single chip can meet this demand. The UK government has again supported this view since April, even if it has not stated so outright.
WSO2 appoints Harry Ault as Chief Executive Officer to drive global growth strategy. WSO2 has announced the appointment of Harry Ault as its new Chief Executive Officer, marking a significant step in the company's growth journey as it expands its presence in the rapidly evolving enterprise technology and AI governance landscape. Ault assumes day-to-day leadership of WSO2 with immediate effect and will work closely with the executive team to strengthen the company's global market position, deepen customer engagement and accelerate innovation across its product portfolio. He brings more than two decades of experience in revenue growth, strategic partnerships, corporate development and global go-to-market leadership. Before joining WSO2, Mr Ault served as Chief Revenue Officer at SambaNova Systems, where he led worldwide sales, marketing, field engineering and customer support operations. Throughout his career, he has played a key role in scaling technology businesses, building high-performance teams and driving growth across international markets. The appointment comes during a period of continued expansion for WSO2. Over the past year, the company has strengthened its leadership team with key appointments in finance and marketing as it advances its vision of trusted AI governance and enterprise transformation. WSO2 is focused on helping organisations manage increasingly complex digital environments by combining API management, integration, identity, engineering and AI-driven platforms within its Agentic Enterprise Fabric (AEF). The platform is designed to provide enterprises with greater control, governance and security as they adopt AI-powered technologies and autonomous agents. Jonas Persson, Chairman of the Board, said Ault's experience in building global teams and driving business growth makes him well suited to lead WSO2 through its next phase of expansion. Harry Ault said WSO2 has built a strong reputation over the past two decades, serving more than 700 enterprise customers worldwide. He added that organisations are increasingly seeking trusted partners that can help them adopt AI technologies without adding complexity, and WSO2 is well positioned to support that transition. Mr Ault succeeds WSO2 Founder Dr Sanjiva Weerawarana, who stepped down as CEO in May 2026. During the transition period, Chief Revenue Officer Devaka Randeniya served as Acting CEO and will continue to work closely with Mr Ault. Founded in 2005, WSO2 provides open-source technology platforms that support API management, integration, identity and AI-driven enterprise applications. The company operates across multiple global markets, with offices in North America, Europe, Asia, the Middle East and Australia, serving enterprises across a wide range of industries.
AMD buys Taalas, the startup etching AI models into silicon. AMD just bought a company whose flagship chip can run exactly one AI model. That is not a design flaw. It is the whole point. The chipmaker announced on August 6 that it has signed a definitive agreement to acquire Taalas, a Toronto startup founded in 2023 with a radical approach to AI hardware. Instead of building general-purpose processors that can load any model, Taalas physically etches a specific model's weights into the silicon itself. Financial terms were not disclosed. A chip that cannot change its mind. Taalas builds what it calls model-specific silicon, and its first product shows how literal that phrase is. The HC1 chip runs Meta's Llama 3.1 8B and nothing else, because the model's parameters are hardwired into the chip during manufacturing. There is no loading weights from external memory, no shuffling data between the processor and storage. The model is the chip. That design attacks the biggest bottleneck in AI inference. Modern GPUs spend an enormous share of their time and energy moving model weights between memory and compute units rather than doing the actual math. By baking the weights into the transistors, Taalas removes that traffic entirely. The chips are fabricated by TSMC on its 6-nanometre process node, a mature and relatively affordable technology compared with the cutting-edge nodes that flagship GPUs demand. The claimed results are striking. Taalas says its hardware can generate more tokens per second per user than Nvidia's H200 and B200 accelerators, and outpaces specialist inference hardware from Groq, SambaNova, and Cerebras, while consuming one tenth of the power. Those are the company's own figures, and independent benchmarks will be worth watching. But even a fraction of that efficiency gain would matter enormously in an industry where electricity has become the scarcest resource. Why AMD wants frozen models. The obvious objection is that AI models change constantly. A chip that runs one model forever sounds like a liability in a field where labs ship new versions every few months. AMD is betting the economics say otherwise. Inference, the work of actually running trained models for users, now dominates AI computing demand. Training happens once. Serving happens billions of times a day, and the workloads are surprisingly stable. A company running a popular model at scale might serve the same version for months. For those customers, a hyper-efficient chip dedicated to that exact model could cut serving costs dramatically, even if the hardware needs replacing when the model does. AMD said it plans to fold Taalas technology into its accelerator roadmap and develop system-level products alongside its Instinct GPUs, EPYC processors, Helios rack-scale platform, and ROCm software stack. The picture that emerges is a portfolio play: flexible GPUs for training and fast-moving workloads, hardwired silicon for the stable, high-volume serving that drains most of the power budget. The inference race tightens. The acquisition lands in the middle of an intensifying contest for the inference market. Nvidia still dominates AI hardware overall, but inference is where challengers see an opening, because the requirements differ from training. Groq built its business on deterministic, low-latency inference chips. Cerebras sells wafer-scale monsters. SambaNova pitches reconfigurable dataflow architecture. Each is chasing the same insight: the hardware that trained the model is not necessarily the best hardware to serve it. AMD has been assembling inference capabilities piece by piece, and Taalas gives it something none of its rivals currently ship, a commercial chip with the model burned in. For Nvidia, the deal is one more sign that the competition has stopped trying to beat its GPUs at their own game and started changing the rules instead. There is also a Canadian angle worth noting. Taalas emerged from Toronto's deep pool of semiconductor and machine learning talent, and its acquisition continues a pattern of US chip giants shopping north of the border for specialised engineering teams. What happens next. The open questions are practical ones. How quickly can AMD turn a startup's first product into something hyperscalers will deploy by the rack? Which models get the hardwired treatment, and who decides? And can the efficiency claims survive contact with independent testing? Answers should arrive over the coming year as AMD integrates the team and slots the technology into its roadmap. If the numbers hold up, the industry may need to rethink one of its base assumptions: that AI hardware has to be flexible. Sometimes the fastest chip is the one that only knows a single trick. For more coverage of AI hardware and the chip industry, visit Mylistingo. Ramo is the editorial voice of Mylistingo - an AI and technology news platform based in The Hague, Netherlands. Covering artificial intelligence, machine learning, robotics, and the future of technology, Ramo delivers accurate, accessible reporting for both general audiences and industry professionals. Every article is fact-checked and written to meet Mylistingo's strict no-fabrication editorial standards.
SambaNova, a leading AI inference company, has appointed Mohsen Moazami as vice chair of global strategy and partnerships. Reporting to CEO Rodrigo Liang, Moazami will drive expansion across global enterprise, government, and sovereign AI markets. The appointment follows SambaNova's Series F financing at an $11 billion valuation. Moazami brings extensive experience from senior roles at Groq, where he served in the Office of the CEO and led international operations through a $20 billion strategic transaction. He also serves as vice chair of DDN, an AI data platform provider. At SambaNova, Moazami will leverage his experience in enterprise, government, and sovereign markets to scale the company's AI infrastructure offerings globally.
SambaNova Brings Mohsen Moazami on as Vice Chair of Global Strategy and Partnerships to Accelerate Global Enterprise and Sovereign AI Expansion By SambaNova August 4, 2026