AgeTech builder, caregiver at heart, and the person behind VEAI LAB. 🌱
My experience as a young carer is why I build tools for caregiving and independent living. I care about the small, everyday moments: preparing for a care transition, remembering what happened today, and making it a little easier to ask for help.
I want these tools to feel warm and approachable, to respect the people using them, and to be honest about what they can and cannot do. They are not tied to any single condition.
I am also a certified project manager (PMP) who writes the code — from browser interfaces to AWS services and automated checks. Below are the projects, experiments and open-source contributions behind that work. 🛠
An offline Python toolkit for inspecting synthetic care observations before they enter analytics or ML pipelines.
- Data quality: deterministic JSON/CSV checks with stable rule IDs and explicit missingness semantics.
- Evaluation: controlled data-quality experiments, including false alarms and missed problems.
- Time and lineage: history replay distinguishes observation, revision and arrival times to prevent later information from entering earlier snapshots.
- Reproducibility: CLI/browser compatibility, synthetic fixtures and CI.
This is data-quality and pipeline engineering evidence. The toolkit does not train or evaluate an ML model, and its examples do not establish clinical performance.
| Project | What to inspect |
|---|---|
| CareQuest | Local-first TypeScript PWA, optional Cognito authentication, backup/restore boundaries, Vitest and Playwright checks. App |
| CareReady | Offline belongings checklist for care transitions. App |
| GutPacer | Bowel and medication records, Lambda/DynamoDB, LINE reminders and explicit observation semantics. In development. |
| ParkinSync | Synthetic health-data research and a deterministic HL7 FHIR R4 evidence path. No clinical inference. |
Merged
- OHDSI DataQualityDashboard #699 —
DEATH.person_iduniqueness checks for CDM 5.3, 5.4 and 5.5. - Home Assistant frontend #54083 — accessible names for analytics consent switches.
- Microduck #241 — fresh camera snapshots through robotctl and console HTTP.
- DuckDB documentation #7309 — recursive Parquet directory inputs, checked against two CLI versions.
Open PRs — status checked on 2026-09-28; links show the current state.
- Microduck #330 — enforcing per-component artifact size budgets and required files.
- stack-chan #702 — deterministic synchronization for gallery samples.
- gitleaks-action #237 — documenting which commits are scanned for each event type.
- Languages: Python, TypeScript and JavaScript.
- Cloud: AWS Lambda, DynamoDB, Cognito, S3/CloudFront and CDK.
- Quality: synthetic fixtures, regression checks, CI and explicit data contracts.
- Delivery: small reviewable changes, documented limitations and traceable implementation evidence.
My solo projects use AI-assisted development. The linked source, checks and review history show what has been implemented and evaluated; they should be read alongside each project's limitations.
🌱 More products: Medication Promise, EchoCare and かんたん家族サポート.
🌐 VEAI LAB. · Engineering and product writing · 📫 Cloud and delivery work