01
Being able to explain why it works
"It works" is not the standard; being able to say why it works is. An incident that recovered without a root cause is, to us, still open.
Careers
Careers
We are looking for people with a durable interest in the technology that sits beneath AI. We are not trying to grow headcount quickly — we want to keep a structure in which every engineer can own design decisions.
What we hold to
What we hold to
02
Owning the unit economics
A technically correct design that does not clear its cost does not survive as a business. We want engineers talking about unit price and utilisation to be ordinary here.
03
Leaving it handover-ready
A system only one person understands is a liability for that person too. Being transferable through code and documentation is part of our definition of done.
Open positions
Open positions
GPU Infrastructure Engineer
Full-time, limited openingsDesign, build and operation of GPU clusters, covering InfiniBand fabric, Slurm / Kubernetes, distributed storage and incident response.
Requirements3+ years operating Linux servers / network or storage design experience / infrastructure-as-code experience
ML Systems Engineer
Full-time, limited openingsDesign and optimisation of inference platforms — throughput, latency and unit cost on vLLM / TensorRT-LLM — plus training platform construction.
RequirementsProfessional PyTorch experience / performance optimisation in training or inference / working knowledge of CUDA and GPU architecture
Solutions Architect
Full-time, limited openingsRequirements definition through architecture, cost modelling and proposal. We are looking for people who can judge on both technical and commercial grounds.
RequirementsCloud or on-premise infrastructure design experience / client-facing experience / grounding in AI and machine learning
Project Manager
Full-time, limited openingsDelivery management for build engagements, vendor coordination and deliverable quality. Experience with engagements involving data centre works is welcome.
Requirements3+ years managing IT infrastructure projects / multi-vendor coordination experience
Working conditions
Working conditions
| Location | Ginza, Chuo-ku, Tokyo (hybrid; typically one to two days on site per week) |
|---|---|
| Hours | Flexitime, core hours 11:00–16:00 |
| Holidays | Two-day weekend, public holidays, year-end period, paid annual leave |
| Compensation | Determined by experience and capability. Full social insurance, books and conference budget, use of internal compute for validation work |
| On-call | Rotational, with an allowance. Consecutive rotations are not assigned. |
Selection process
Selection process
Application review→Informal conversation→Technical interviews (1–2)→Final interview→Offer
The informal conversation is not part of the assessment. It is mostly your questions about the business and the architecture. You are welcome to book one without intending to apply.
Applications and enquiries
Please send a CV in any format to our recruitment address. A GitHub profile or technical writing is equally acceptable in place of a formal CV.