Customer Engineer at Google Cloud Japan, working with financial services enterprises, and leading development of an internal Google application. Recent customer work keeps landing in the same few places: agentic applications, harness engineering for enterprise agents, semantic knowledge catalogs, and landing zone design on Google Cloud.

Before that, several years at a Japanese systems integrator across most of the roles a project has: systems engineer, application developer, project manager, engineering manager, and tech lead on a new service. The infrastructure work was Windows — Active Directory, Exchange, client networks — and the new-service work leaned on Lean Startup and Agile practices.

I have just finished a Master of Computer Science at UIUC — coursework complete, graduating Summer 2026 with the diploma pending. My capstone research was MASReasonCache, a reasoning cache for multi-agent systems: it mines past agent execution logs with association-rule learning, then injects the resulting rules back into later runs as a shared semantic cache, so agents stop re-deriving conclusions the system has already reached. Built on Google's Agent Development Kit with MongoDB and ChromaDB, and evaluated on a four-agent travel-planning system, it cut token usage by as much as 78% on the most complex reasoning tasks.

Day to day that means TypeScript and Python, with Next.js, React and Flutter on the front end, Google ADK and LangChain for agents, and Google Cloud underneath. Claude Code and Antigravity have become part of how I build rather than things I occasionally reach for.

This blog is where I write about that work: web technology, AI agents, Google Cloud, Agile, and project management.

This blog is called overOptimized(). It replaced an earlier one called isOverwhelmed?, which is more or less the trajectory.

The views expressed on this blog are mine alone and not those of my employer.