Work

For a major Japanese shopping-mall operator, I led a digital kiosk that helps shoppers find stores and services, and call a human operator by voice when they need help.

What I contributed

I elicited requirements, designed the solution, ran the project, and led development — coding myself while directing three engineers. I ran weekly meetings with the client in Japanese, stayed on call for support, and owned uptime, security, versioning, and rollout.

Stack

Python backend, Kotlin Android app on the tablet, React admin panel, AWS (EC2, Route53), WebRTC voice calling, LLM/voice APIs, PostgreSQL.

Outcome

The AI customer kiosk is now live in about five locations, with a national rollout planned for 162 locations across Japan.

For a hospital-supply client, I led a system that helps hospitals manage inventory, resources, tasks, and equipment repairs — including a multi-step repair workflow from intake through vendor quotes and completion.

What I contributed

I estimated cost and capacity, managed the delivery schedule, led development (coding plus directing five engineers), ran code reviews and engineering policies, and coordinated with the product manager, UI/UX designer, and technical designer.

Stack

Python backend, React frontend, AWS (including ECS), large language models where they helped operators, Docker, CI/CD.

Outcome

We delivered a production-ready operations suite the hospital can run day to day, with clear ownership of quality and schedule across a five-person engineering team.

For a leading Japanese university, I was Tech Lead for the AI and API behind a research showcase chatbot — answering questions about researchers and projects from a curated knowledge base.

What I contributed

I elicited requirements, ran AI prototyping (data ingestion, prompt experimentation, RAG setup and optimisation), designed the API, developed and deployed to AWS (EC2 and ECS), managed cost and schedule for the AI/API slice, oversaw testing, and joined weekly meetings with the university and partner companies in Japanese.

Stack

Python, Weaviate vector database, PostgreSQL, Nginx, LLM APIs, AWS EC2/ECS.

Outcome

Visitors can explore research in natural language; the AI/API path met functional and non-functional requirements (including scalability) under joint stakeholder review.

For a Japanese steel-industry client, I led development on Archaic’s document chatbot so staff can search internal procedures in natural language and jump to the referenced source documents.

What I contributed

I led feature work (coding plus one engineer), including semantic chunking of long procedures and linking answers to referenced documents, then deployed to AWS and checked that answers retrieved the right source procedures.

Stack

Python, React, AWS EC2, Weaviate vector database, PostgreSQL, Nginx, LLM APIs.

Outcome

Operators can find the right procedure without knowing the filename, with answers grounded in links back into their document set.

For a major Japanese printing and media company, I adapted Archaic’s Chatbot Anything product so agents and systems can look up responses to common customer inquiries.

What I contributed

I tailored the product to their requirements, deployed to AWS, and checked answer quality against their FAQ and support content before handoff.

Stack

Python, React, AWS EC2, Weaviate, PostgreSQL, Nginx, LLM APIs.

Outcome

A production inquiry assistant aligned to their FAQ and support content, with QA gates before handoff.

For a Japanese insurer’s driving app, I led an API that lets company clients schedule and receive AI-generated reports on drivers.

What I contributed

I elicited requirements, designed the solution, led development (coding plus directing two engineers), ran code reviews and the schedule, and owned quality — functional requirements plus non-functional ones — including overseeing testing before release.

Stack

Python, PostgreSQL, AWS (EC2, Cognito), OAuth 2.0 / SSO.

Outcome

B2B clients can request driver reports through a documented API with authentication suitable for enterprise integration.

For a major Japanese residential developer’s mobile app, I delivered systems that turn logged health, nutrition, and exercise data into regular health reports and quarterly exercise plans — combining ML models and rule-based advice.

What I contributed

I elicited requirements, trained and fine-tuned models on client data, designed and built the services, took over exercise-recommendation work after its POC, investigated production issues, optimised performance (including Cython where it helped), deployed and maintained on AWS, and met the client regularly in Japanese. Related dashboard API work for the same app covered health logs, step recommendations, and health scores.

Stack

Python, PyTorch, AWS (EC2, VPC peering, QuickSight), Cython for hot paths where needed.

Outcome

Residents receive scheduled health and exercise guidance from data they already log — with models and rules I could iterate under client review.

For a global car manufacturer and a telecoms partner, I helped build a product that lets engineers browse and search technical diagrams of mechanical parts — including text search powered by OCR and retrieval over diagram content.

What I contributed

I designed and developed the AI features (OCR over diagrams, text search, and document-backed answers), stood up AWS environments, and worked with the partner to integrate and deploy — Nginx reverse proxy, domains, firewall and security rules, and stability / operations checks.

Stack

Python, React, AWS (EC2, Route53), Weaviate, PostgreSQL, Nginx, LLM APIs.

Outcome

Teams can find parts and diagram detail by meaning, not only by filename — with the stack running in the partner’s environment.

For a major Japanese printing and media company, I led adaptation of Archaic’s AI Concierge — real-time voice conversation with AI — so their customer-service teams can practise realistic calls.

What I contributed

I gathered requirements, designed the solution, estimated cost, managed the schedule, led development (coding plus one engineer), reviewed code, and folded in client feedback.

Stack

Python, React, AWS, LLM and voice APIs.

Outcome

Trainers get a voice practice product tailored to the client’s scenarios, built on a proven Archaic platform rather than a greenfield rewrite.

For a major Japanese residential developer, I led the software side of a smart-camera system that watches for security and health events in apartment buildings — for example someone entering without a key, or entering a bathroom and not coming out.

What I contributed

I owned APIs, infrastructure, documentation, QuickSight dashboards, incident investigation, cost reduction, and integration with partner APIs. I led two developers and held regular meetings with the client and stakeholders in Japanese. Related work for the same client included health-report and exercise-recommendation ML services and a dashboard API for their residential mobile app.

Stack

Python, AWS (EC2, Lambda, RDS, VPC peering, QuickSight), image recognition and, for adjacent products, PyTorch models.

Outcome

Operators gained monitored zones and alerts they could act on; we kept improving the system against live client needs rather than a one-shot delivery.

During the pandemic I invented Social Distance AI inside Archaic — computer vision that detects people and helps organisations monitor distancing rules.

What I contributed

I designed and built the detection pipeline, iterated with real deployments (including an education provider and a development agency), and supported production use. I am credited as inventor on Archaic’s patent application for the technology — not only named on the filing, but recognised as the inventor of the system.

Stack

Python, person-detection models, on-prem and cloud camera deployments.

Outcome

A productised capability Archaic could sell and deploy quickly when clients needed distancing tooling — and a durable IP claim for the company.

Kanpeki Study started as a tool for my own Japanese study and grew into a paid product I still run alongside Archaic.

What I contributed

Architecture and implementation, hiring and managing a UX/UI designer, integrating designs, performance and security work, marketing, and continuous iteration from user feedback.

Stack

PHP / Laravel, Vue / Blade / AlpineJS, PostgreSQL, MeCab, LLM APIs, Docker, GitLab CI/CD, AWS (Lambda, RDS, DynamoDB, Route53, CDN, S3, WAF).

Outcome

300+ registered users and over $1000 in sales on a full AWS stack I still operate — an independent product that proves I can own product as well as client delivery.

Archaic is an AI company in Tokyo. As Tech Lead (permanent since November 2019) I own design, architecture, development, and team management across concurrent client programmes.

What I specialise in here

Turning AI prototypes into systems clients can operate: chatbots over company knowledge, voice experiences on tablets and phones, reporting APIs for insurance and mobility apps, and vision systems for buildings and exams. I work in Japanese with stakeholders every week.

Typical stack

Python, JavaScript / React, SQL / PostgreSQL, Docker, AWS, retrieval-augmented generation and LLM APIs, Weaviate, Nginx, CI/CD.

Key client programmes from this role are listed on this page — mall kiosk, hospital operations, research and enterprise chatbots, inquiry and driver-report APIs, health ML, diagram search, voice trainer, smart camera, and Social Distance AI — plus remote UK frontend engagements.

For a UK-based data-analytics company, I worked remotely as a frontend engineer converting legacy vanilla JavaScript widgets to modern React and TypeScript — from simple tables through complex charts.

What I contributed

I migrated widgets one by one, checking for bugs and performance regressions so behaviour stayed equivalent under the new stack. Delivery was async and overseas — the same remote pattern I use for Tokyo-based freelance work with international clients.

Stack

JavaScript, React, TypeScript.

Outcome

A growing set of widgets on a maintainable React/TypeScript base, shipped remotely to a UK team.

For a UK-based real-estate firm, I built the frontend of a dashboard that visualises property data from APIs according to Figma designs — solo, remote delivery.

What I contributed

I implemented the UI to the client’s requirements and designs, integrating API data into clear charts and views. Another proof point for overseas async frontend work from Tokyo.

Stack

JavaScript, React, TypeScript.

Outcome

A Figma-faithful dashboard the client could use to explore real-estate data without a local engineer on site.

As contractor Lead Software Engineer at PORTERS in Tokyo, I worked on HR Business Cloud — matching recruiters with applicants.

What I contributed

Feature design and development, bug fixes, technical documentation, mentoring juniors, and code reviews in a Scrum team. I often picked up late emergency work when delivery risk appeared, and we still shipped on schedule.

Stack

Java, PHP, JavaScript, MySQL, Jenkins, Docker.

As permanent Software Engineer at GE (General Electric) in the United Kingdom, I owned development and maintenance of test software for the Subsea Electronics Module 2000 (SEM2K).

What I contributed

Ensuring units met quality standards, managing test equipment and software, investigating failures, and improving the Java test toolchain.

Stack

Java.

I spent a year as a Computer Scientist at CERN in Switzerland, supporting the CMS experiment.

What I contributed

• XDAQ — features and fixes for CERN’s internal data-acquisition library (C++).
• Oracle TimesTen — feasibility research for in-memory database use in performance-critical CMS workflows.
• Heartbeat — a server-farm health visibility UI (Adobe Flex / ActionScript).