About me
I'm Jesse, a product designer and engineer currently based in Brisbane, Australia. I turn how a business really works into software people keep using: product design and frontend engineering in one head, ten years of B2B, and now LLM features with a person at every decision that matters.
Most of my work starts somewhere that runs on paper, spreadsheets or habit. I work out with the people there what should change, then design and build the system and stay until it's actually used. I'm currently finishing a Master of IT at QUT in Brisbane, and I work in Chinese (native), English, Japanese and Korean — which matters more than it sounds, because most of this work is translation of one kind or another.
What I do
Digital and AI transformation in legacy operations. Helping a paper-run welfare operator onto digital workflows at DEN, and building LLM document checking for KOD, with requirements worked out in Japanese, directly with the people who'll use the system, and a clear line between what code decides, what a model drafts and what a person signs off.
Product design and frontend engineering. A redesigned booking flow at WAmazing that cut peak-season inquiries from over 200 a week to around four, frontend leadership on enterprise HR SaaS at Money Forward, and LingoBun, my own AI product, built and tested like production software.
Design infrastructure in low-maturity teams. As the first designer at Zerospec, I set up the process and the design system for its GoNOW product, so that design held together after the first screens shipped.
Where I come from
DEN Inc. — social welfare, eight years. I helped a 20-person facility running on paper become a multi-facility business of nearly 100 employees on digital workflows. Employee productivity and profit both moved, and several thousand dollars of monthly revenue now runs without anyone touching it. Social welfare is among the least digitised sectors in Japan — none of this was the default outcome.
Money Forward Inc. — HR Cloud, frontend lead. I led frontend development on an enterprise HR SaaS product and held the team to one rule: a change to the code had to be justified by a better experience for the user, never the other way around.
WAmazing Inc. — ski-pass booking, design and delivery. Legacy code and a peak season with no room for downtime meant every change had to be priced in engineering risk first. I did that pricing, and a redesigned booking flow cut peak-season customer inquiries from over 200 a week to around four.
Zerospec Inc. — energy, first designer on the team. On its GoNOW product, I introduced a prototype-driven, implementation-aligned process and the design system to hold it together, and moved a company where no one took design seriously into one that plans around it.
What I'm building now
LLM document checking for KOD — client work. I built a pipeline that reviews long industrial manuals for a Japanese manufacturer: text and vector-drawn warning-symbol extraction from PDF, a rule engine the client can maintain themselves, and agent orchestration with LangGraph. Requirements were scoped and negotiated in Japanese, directly with the client.
AI-drafted compliance records at DEN — pilot. A system, now in pilot, to replace spreadsheet workflows across multiple welfare sites. An LLM drafts regulatory activity records under a three-tier model — deterministic checks in code, LLM self-assessment, and mandatory human sign-off — so nothing is issued that a person hasn't confirmed.
Evaluating LLM security triage — QUT capstone. An evaluation harness over 146 labelled attacks across 14 techniques on a Wazuh SIEM. The most useful findings weren't about the model: logging design explained most detection gaps, and a lenient scoring rule had been inflating accuracy by more than 20 points.
LingoBun — my own AI product. A vocabulary app built and hardened like production software: rate limits and cost caps on AI and text-to-speech calls, validated configuration, and 180 automated tests, shaped by rounds of user testing.
How I think
- Designing the system, not the screen — why design effort belongs in the infrastructure behind the assets.
- From requirements to assumptions — building to learn instead of specifying up front.
- How Claude, ChatGPT and Google AI Mode handle the Guidelines for Human-AI Interaction — where AI product design still owes its users an answer.
- AI's productivity gains aren't evenly shared — a designer's view of who actually benefits.
What I'm looking for
Senior IC roles where design, engineering and AI meet inside real operations — design or UX engineer, AI product engineer, technical director, forward-deployed engineer — in Tokyo, from early 2027. I'm also always glad to talk with teams bringing digital and AI tools into Japanese businesses.
Find me on LinkedIn or GitHub, or email jesseweijapan@gmail.com.