Refolk
Open nowEngineering

Japan Chief Architect

SambaNova · Tokyo Prefecture, Japan

Location
Tokyo Prefecture, Japan
Employment
Full time
Level
Executive
Posted
20 November

About this role

SambaNova is a leader in next-generation AI infrastructure, delivering a full-stack inference platform for customers worldwide. At the core of SambaNova's technology is the RDU (Reconfigurable Dataflow Unit) - a chip built on a dataflow architecture rather than the traditional GPU model. Its decode performance is especially strong for agentic workloads like multi-turn agents, code generation, and long-running applications. RDUs are packaged into SambaRack, rack-scale hardware that lets customers deploy state-of-the-art models with better performance, greater energy efficiency, and faster time to value.

We are building a next-generation AI inference accelerator and the rack-scale system around it, in Japan, on a leading-edge domestic 2nm process. The work is funded through a multi-year national programme to establish sovereign AI infrastructure, and runs from architecture definition through test silicon, full chip tape-out, and deployment of production racks.

We are looking for the architect who will define the chip and technically lead the team that builds it, working with the US HQ engineering team. You will decide how compute, memory and interconnect are balanced for large language model inference, and what the hardware exposes to the compiler and runtime. You understand the software stack as well as the silicon, and you stay with the architecture through RTL, verification and physical implementation until it works in silicon.

What you will own

  • The chip architecture and microarchitecture specification for a reconfigurable dataflow accelerator: tile architecture, on-chip network, memory hierarchy, and the execution model that maps AI workloads onto them.
  • The hardware and software interface: the programming model, control interfaces, and the assumptions the compiler and runtime are allowed to make about the hardware.
  • Technical leadership of the implementation, from architecture through RTL, verification and physical design, to tape-out and silicon correlation.

Responsibilities

Architecture definition

  • Define the chip and system architecture against inference performance requirements: tokens per second per user, aggregate throughput, and tokens per second per watt.
  • Use performance models and architectural simulations to decide trade-offs, and keep them calibrated against measured silicon as it becomes available.
  • Make the balance calls across compute capability, memory bandwidth and interconnect bandwidth, and defend each one with data.

Hardware and software co-design

  • Work directly with compiler, runtime and model optimization engineers so that the architecture is one the software stack can actually exploit.
  • Own the interfaces that software depends on: mapping, scheduling, memory allocation, synchronization, power, and scale-up and scale-out networks.
  • Analyse real LLM workloads (attention, KV cache, quantized and low-precision arithmetic, batching) on the architecture, and feed the results back into both the hardware and the software.

Technical leadership of implementation

  • Lead the team technically from specification to tape-out: microarchitecture specs, RTL design reviews, block interfaces, clocking, reset and power domains.
  • Set verification goals with the verification lead, and ensure performance correlation between the architectural model and the RTL.
  • Assess physical feasibility early and keep the design within it: area and timing budgets, floorplan partitioning, power, thermal limits, and die and package size.
  • Lead technical design reviews at each integration milestone, and give the technical sign-off for tape-out.
  • Support silicon bring-up, correlate measured silicon against the models, and carry the lessons into the next generation.
  • Mentor architects and designers in Japan, and act as the technical counterpart to US HQ architecture so that the two teams build one architecture.
  • Present the architecture in technical reviews with programme sponsors, the foundry, and IP and system partners.

Minimum qualifications

  • 20 or more years in silicon architecture and design, including architecture ownership on complex SoCs or accelerators.
  • Served as an architect on at least three chips taken from concept through tape-out to working production silicon.
  • Background in GPU, AI accelerator, or high-performance CPU and SoC architecture.
  • Working command of the hardware and software boundary: how compilers, runtimes and kernels target an accelerator, and which architectural choices make a target easy or impossible to use well.
  • Practical understanding of implementation (microarchitecture and RTL, verification, floorplanning, timing convergence, power delivery and packaging), deep enough to know when an architectural idea will not survive them.
  • Experience with architectural performance modelling and quantitative power, performance and area analysis.
  • Experience technically leading a team of architects and designers through tape-out.
  • Professional working English; fluent Japanese strongly preferred. The role works with the team and partners in Japan and with engineering teams outside Japan.

Preferred qualifications

  • Hands-on RTL design experience in SystemVerilog or Verilog on high-performance blocks.
  • Experience with GPU programming models and software stacks (for example, CUDA, Triton, or ML compilers such as MLIR or XLA).
  • Dataflow, spatial or reconfigurable architectures, and the compiler models that drive them.
  • Large language model inference at scale: attention and KV cache behaviour, quantization and low-precision arithmetic, long context, and batching strategy.
  • High-bandwidth memory subsystem architecture, and multi-tier memory hierarchies.
  • Chiplet, die-to-die interface, and advanced packaging or interposer architecture.
  • First-use experience on a leading-edge node where the process design kit, libraries and interface IP were still maturing during the project.
  • Advanced degree in electrical engineering, computer engineering or computer science.

Base Salary Range:

Base Pay Range¥32,670,000-¥39,900,000 JPY

Submission Guidelines
Please note that in order to be considered an applicant for any position at SambaNova Systems, you must submit an application form for each position for which you believe you are qualified.

EEO Policy
SambaNova Systems is an Equal Opportunity/Affirmative Action Employer. All qualified applicants will receive consideration for employment without regard basis of age (40 and over), color, disability, gender identity, genetic information, marital status, military or veteran status, national origin/ancestry, race, religion, creed, sex (including pregnancy, childbirth, breastfeeding), sexual orientation, and any other applicable status protected by federal, state, or local laws.

Benefits Summary for US-Based, Full-Time Employment Positions
SambaNova offers a competitive total rewards package, including the base salary, plus equity and benefits. We cover 95% premium coverage for employee medical insurance, and 77% premium coverage for dependents and offer a Health Savings Account (HSA) with employer contribution. We also offer Dental, Vision, Short/Long term Disability, Basic Life, Voluntary Life, and AD&D insurance plans in addition to Flexible Spending Account (FSA) options like Health Care, Limited Purpose, and Dependent Care. Our library of well-being benefits available to you and your dependents includes a full subscription to Headspace, Gympass+ membership with access to physical gyms, One Medical membership, counseling services with an Employee Assistance Program, and much more.

As published by SambaNova. Applications are handled on their site.

About SambaNova

At SambaNova, we're empowering developers, enterprises, governments, and data centers to unlock their full AI potential. Our full-stack infrastructure, from chips to models, enables lightning-fast performance, low power consumption, and high-efficiency computing.

All 75 openings at SambaNova

One click, then it is written

Apply to SambaNova with a resume written for this role.

Queue Japan Chief Architect and I read the posting, rewrite your resume against it, draft the cover letter, and score the fit. Then you press send, or press one button and I fill in SambaNova’s form for you.

  1. 01Drop your resume

    A PDF or a LinkedIn URL. About a minute, once.

  2. 02I rank the openings

    Every weekday morning, the live catalog scored against your history. Up to 20 worth your time, not two hundred links.

  3. 03Each one is written up

    Resume rewritten for the posting, a cover letter, a fit score. Press send, or let me fill in the form.

  • 25 sent a week, free
  • No card
  • Nothing sent until you say so

More roles at SambaNova

See all

Similar roles elsewhere

See more

Put this to work

Paste your career in once. Every application after that is written for you.

Drop a resume or a LinkedIn URL. I rank the live openings against it, rewrite the resume and write a cover letter for the best of them, and fill in the employer's form when you press the button. You read, you decide what goes out.

  1. 01Drop your resume

    A PDF or a LinkedIn URL. About a minute, once.

  2. 02I rank the openings

    Every weekday morning, the live catalog scored against your history. Up to 20 worth your time, not two hundred links.

  3. 03Each one is written up

    Resume rewritten for the posting, a cover letter, a fit score. Press send, or let me fill in the form.

  • New matches ranked and written before you are up.
  • Every bullet stays inside what your history supports. Nothing invented.
  • Queued, submitted, interviewing, offer: one screen, not a spreadsheet.

Free, no card. Up to 25 applications sent a week, and nothing is sent until you say so.

Listed from the job board SambaNova publishes. Refolk is not the employer and does not handle their hiring. Applications go to SambaNova directly.

Hiring at SambaNova? Post this role to the Refolk talent pool for $299, once. No commission, refunded if I find nobody. Post a role for $299