I build evidence-driven software products end to end — from the research behind a formula to the shipped, deployed product a real person uses.
The throughline across everything below: dig into the actual data or process a product needs to serve, disclose what's rigorous versus what's a calibrated estimate, and match the architecture to the real constraint instead of defaulting to whatever's trendiest. Based in Malaysia, building under the Aufthority label.
Core strengths: product development · data governance & evidence auditing · process/compliance design · applied research & population-calibrated analysis · access-control & schema design
Three areas, one method.
Screening and tracking apps calibrated to Malaysian/Southeast Asian population data instead of defaulting to Western norms.
- CerminDiri — metabolic & cardiovascular risk screener: FINDRISC, INTERHEART NL-IHRS, Framingham (BMI version), Asia-Pacific BMI cutoffs
- MyTDEE — calorie calculator built on a Malaysian-derived BMR equation (Ismail et al., 1998) instead of Western defaults
- MakanLah — photo-first meal diary for dietitian review; no calorie counting, on-device storage only
- NeuroHabit — habit tracker where every habit is individually weighted against a literature review, not just brainstormed
Process and compliance systems built for real institutions with real constraints — no dedicated dev support, real regulatory requirements.
- Kompas Semester — course companion with a proper Postgres schema and Row-Level Security, built after trending coursework results against final exam performance surfaced a measured mastery gap
- KPJU-SOP CPD Tracker — live compliance dashboard mapping a CPD programme's full approval-to-close-out lifecycle, revenue tracking included
- KPJU CPD Bot — Telegram FAQ/deadline bot backed entirely by a Google Sheet; zero database, zero redeploy to update content
- Evidence-tiered, not just built. Every formula, threshold, and habit weighting is labeled Measured / Proxy / Design-calibration and disclosed in the product itself — not asserted.
- Gaps disclosed, not hidden. Where no population-specific dataset exists, the product says so rather than quietly substituting a Western default and presenting it as equivalent.
- Architecture matched to the actual constraint. Zero-backend, single-file static apps by default; a real Postgres backend with Row-Level Security only when the product genuinely needs persistent multi-user data — a deliberate choice per project, not a default.
- Built for the people who'll actually use it. Bilingual (EN/BM) across the health tool suite. PDPA-aware consent flows where personal data is collected. Culturally specific content where it matters — prayer and dhikr as literature-backed habit categories, not generic wellness copy with a translation layer.
A consolidated ledger tracking the evidence tier — measured, proxy, or design-calibration — behind every formula and threshold across the tool suite: aufthority/evidence-ledger
HTML CSS JavaScript PostgreSQL Supabase Row-Level Security Vercel Firebase PWA / Service Workers Chart.js Google Sheets API / CSV pipelines