Hands-on restaurant operations background meets AI × automation. Solo practice in Tokyo, building production systems for restaurants and small businesses.
Get in touchI'm Masahito Hebiishi (蛇石 真民). I spent eight years on the floor at Skylark Holdings, one of the largest restaurant groups in Japan, before transitioning to engineering. Today I design, build, and operate AI and automation systems for restaurants and small businesses — alone, end to end.
Floor work at Skylark Holdings: operations, numbers, staff training, daily store running.
Moved to engineering to fix the inefficiencies I saw daily. React, Next.js, FastAPI, Cloud Run.
Notion, Google Apps Script, AI, and web apps. Five products running in production. Based in Tokyo.
Get restaurants visible in ChatGPT, Google AI Overview, Perplexity, and — in Japan — LINE Agent i. Structured data, content engineering, monthly measurement.
LINE AI Chatbot (β) setup, Q&A engineering, preparation for the LINE OA AI Mode launching summer 2026. Specific to the Japanese market where LINE dominates.
Database design, relations, automated sync with Sheets/Forms/Gmail. 4,200+ Notion page sync flows shipped.
Claude, OpenAI, and Gemini APIs for summarization, classification, report generation. RAG-based internal knowledge search (8,200+ chunks, 87.5% accuracy in production).
React / Next.js / FastAPI / Cloud Run. Internal tools to consumer SaaS. Currently operating five production apps.
2,400+ national exam questions with AI explanations, AI meal-plan generator, RAG-based nutrition chat. Three-tier subscription.
Next.jsPrismaPostgreSQLClaude APIGemini APIStripe
Mental-health focused anonymous community. Crisis keyword detection and SOS button built in. iOS / Android.
ReactFirebaseCloud RunCapacitor
Test how ChatGPT, Gemini, Claude, and Perplexity describe a brand. Integrated reports with competitive deltas and action items.
ReactExpressPostgreSQLStripe
Auto-collected prospect lists via Google Places API and scraping. Team management and Stripe billing built in.
StreamlitPostgreSQLCloud RunStripe
Natural language → SQL → charts → AI interpretation, end to end. Built for small restaurants who can't afford an analyst.
FastAPIReactEChartsClaude API
Design, implementation, and operation are the same person. No broken telephone between requirements and code, no runbooks written by people who never ran the system.
Floor-trained intuition shapes the design: I know what "the staff won't use this" looks like before code is written.
Decision speed is the moat. A LINE message at noon can become a deployment by evening.
If you're working on restaurant tech, AI-in-messenger-app strategy, or AI automation for SMBs, I'd be happy to talk.