ENTRY 02 · BUILD · STATUS: ACTIVE ▮
I build what I need.
I have an IT degree in networks and security and a habit I can't shake: when a process annoys me, I build the tool. Everything below is mine: designed, built and run by me, mostly nights and weekends. Work-adjacent tools are shown sanitized; client data never appears here.
Compliance Management System
The internal operating system I run the whole compliance function on. Local-first, on my own hardware.
Kravios
A compliance operating platform for MAS-regulated fund managers. Fictional firms, real deadlines.
AML/CFT training platform
In-house annual AML/CFT and proliferation-financing training, with a quiz the client can't cheat.
ReelForge
A node-graph editor for 9:16 video, with a human approval gate after every generative step.
Memory Bridge
A privacy-first family oral-history app: guided recording sessions between children and aging parents.
HRDex
A Singapore-first workforce operating system: HR, leave, claims, payroll and approvals, built around who is allowed to see what.
EasyShifts
Multi-tenant shift scheduling and payroll, built for and used by a boutique fitness studio in Singapore.
SplitLah
Bill-splitting for trips, shared rent, and mahjong nights. Multi-currency, and the maths is tested.
麻將 Pro
Singapore mahjong in the browser: four AI difficulty tiers, online rooms, four rulesets.
3D Explore
A browser fantasy exploration game, and a repeatable pipeline for turning free 3D assets into a world.
Contractor marketing site
A rebuilt marketing site for a Singapore M&E and air-conditioning contractor, built for credibility, not decoration.
ALSO IN THE REGISTER: IBKR ANALYST · MECHAMORI, A UNITY RPG BUILT BY ORCHESTRATED AGENTS · TATAMI, STRAVA FOR GRAPPLERS · MERIDIAN HR · A NUTRITION TRACKER · MAHJONG SETTLEMENT ENGINES · A FLOOR-PLAN DESIGNER · A BROWSER JRPG. SOME SHIPPED, SOME PAUSED. ALL LOGGED HONESTLY.
◇ PRINCIPLE
AI that stays inside the walls.
Anything sensitive runs on models I host myself. Client data does not go to other people's clouds, not because a policy says so, but because I wrote the policy. Every build here was made this way, and most of them were built with agents doing the work under supervision.
I am not going to write a tutorial. If you want to know how any of this is put together, or you are trying to get agents doing real work inside a regulated business, ask me directly.
◇ INFRASTRUCTURE
The always-on layer.
Two systems run whether I'm at the desk or not. A personal agent lives on my Raspberry Pi and works around the clock: it watches my services, runs the scheduled jobs, and messages me when something actually needs a human.
Everything I learn ends up in my AI second brain: an Obsidian vault where notes on compliance, code and training link into one connected memory, kept current by an agent so nothing gets lost. The published slice of it is in the second brain.