Founded in 2021 in Kyoto. We build production AI, SaaS, and integration software for teams that want the engineering done once, properly.
[ A studio building intelligence as infrastructure ]
"Three commitments we make on every project — measured, not promised."
Every model is evaluated, red-teamed, and benchmarked before it touches production.
We don't ship notebooks. We ship services with SLOs, monitoring, and on-call rotations.
Success is measured in retained users and revenue — not model benchmarks.
Seven steps, repeated every engagement. No agency theater, no surprise reveals. Each step has its own artifact — so the next person who touches the codebase has a clear story.
"We don't pitch a process. We rehearse the same one until it ships every time."
Interviews, system audits, and constraints — we learn what's true before we propose what's possible.
Architecture diagrams, decision records, and a scoped problem statement everyone signs off on.
Information design and UI — pressure-tested with your stakeholders before a line of production code.
Weekly demos, trunk-based development, tests as we go. No surprise reveal at the end of the project.
Eval suites, performance budgets, and stakeholder sign-off against the success criteria set in step 02.
Staged rollout, observability, runbooks, and a handover the team taking over actually wants.
Quarterly architecture reviews, performance work, and the next wave — under retainer or a fresh sprint.
Three leaders. Three disciplines. One obsession: shipping AI that earns its place in production. Meet the operators architecting ReimeiTech's edge.
15+ years shipping ML at Google Brain and Tokyo Financial Hub. Kenji believes intelligence is not a product — it's a posture. He leads ReimeiTech's thesis on building AI that outlives the companies that ship it.
We're not chasing AGI. We're building the infrastructure that runs under it.
A distributed-systems engineer with deep roots in ML infrastructure, Satomune architects every system ReimeiTech deploys — from token budgets to evaluation harnesses. He ensures reproducibility is never sacrificed for velocity.
If you can't measure alignment, you haven't shipped it.
