SAI
DeNA · Internal AI platform · Product Owner (0→1) → Fullstack Engineer
- 800
- MAU in 3 months
- ≈ 2/3
- of all employees
- 40,000h
- saved per year
- $1.8M
- infra cost cut /yr
DeNA had no first-party generative AI platform. Employees across game, health, sports, and live-streaming businesses needed safe, fast access to LLMs as part of their daily work. I led the product from a blank Figma to company-wide adoption, then went back into the codebase to make it cheap to keep alive.
Hit 800 MAU (≈2/3 of all employees) within 3 months. Estimated ~40,000 hours/year saved across the organisation. After product-fit, I refactored the architecture as a fullstack engineer, cutting infrastructure costs by $1.8M/year.
Logion
Personal project · Acquired May 2025 · Solo Founder · Builder
- 3 mo
- build → acquisition
- 0 → 1
- shipped solo-led
Non-technical business users wanted to ship custom LLM workflows without waiting on engineering. I built Logion as a UI-first generative AI app with native integrations into Dify and similar platforms — and ran the company at AI-native development velocity.
Built and shipped to acquisition in 3 months. Designed for non-technical users to deploy custom LLM workflows without engineering resources.
Enterprise LLM Fine-tuning
ELYZA · UTokyo Matsuo Lab spinout · Lead Machine Learning Engineer
- ~30%
- efficiency gain
- 500–700h
- saved per month
In 2023 — before most companies had touched generative AI — I led enterprise fine-tuning projects for one of Japan's earliest LLM startups. The work was equal parts research, customer engineering, and convincing organisations that LLMs were ready for the inside of their workflows.
Delivered ~30% efficiency improvement and 500–700 hours/month saved for a major enterprise client.
Virufy Japan
Stanford COVID-19 Innovation Lab · Country Lead (Japan)
- 30+
- team built in 3 months
- 12
- PCR labs coordinated
Virufy was a Stanford-led project building a smartphone app that detects COVID-19 from cough sounds with AI. I owned Japan operations: clinical data collection at scale across hospitals, PCR labs, and waiting hotels — across regulatory, logistical, and international boundaries.
Built a 30+ member team from scratch in 3 months. Coordinated with 12 PCR laboratories, hospitals, and waiting hotels for clinical data collection. Passed ethics committee reviews. Collaborated with international teams in the US and Colombia.
