Work
Built, not promised.
Shipped work, described honestly. Clients are not named.
01 — Money changing & remittance XChange
A digital platform for a licensed money changing and remittance business.
The problemA licensed money services business needed a digital platform, compliant by design — KYC, screening, reporting, encryption, audit — every requirement documented against the regulator's checklist.
What we builtRust API, Next.js front end, branch agents driving money counters and receipt printers. Field-level encryption throughout; WireGuard-gated access — no VPN, no app.
The outcomeIn production on Alibaba Cloud, Kuala Lumpur — with the full compliance package prepared for submission. One account per client, same hardening — repeatable.
- Pillar
- Software · Cloud
- Stack
- Rust · Next.js · MySQL
- Status
- In production
02 — CRM & business operations CRM & invoicing platform
A business-operations platform: CRM at the core, with finance, marketing and admin dashboards.
The problemCustomers, leads, payments, invoices — one system instead of several, never on the open internet.
What we built.NET 8 API, Next.js front end, invoices from templates to PDF. Per-user WireGuard peers; servers on private addresses only — every request attributable to a user.
The outcomeDeployed for its first client under the no-VPN-no-app model; the canonical deploy configs are ready for the next.
- Pillar
- Software · Cloud
- Stack
- .NET 8 · Next.js · MySQL
- Status
- Deployed
03 — Custom AI model Coin-Toss
A stock-trading ecosystem for Bursa Malaysia — data organized, analyzed and modelled, with AI-assisted picks.
The problemInvestors on Bursa Malaysia piece their trading picture together by hand — scattered data, spreadsheets, gut feel. The brief: one system that organizes, analyzes and models stock trading, and assists the trade.
What we builtA whole ecosystem for stock trading — data organized from multiple sources, analysis and modelling on top, surfacing ranked picks, watchlists and a trade simulator.
The outcomeTraining runs on spot instances for about a dollar per run, and the walk-forward simulator has logged nearly 48,000 backtest trades.
- Pillar
- AI model
- Stack
- Python · PyTorch
- Status
- In use
Your project belongs on this page.
One paragraph is enough — we'll propose a shape.
Contact us