James Kyle De Leon

The attendance system every employee clocks into — 15 biometric terminals, 500+ staff — ran on an on-premises server and had to move to Google Cloud without interrupting attendance capture. I planned and executed the migration of the ZKBioTime server to Google Cloud and re-pointed all 15 terminals to the new environment, delivered as an assigned project alongside daily operations, with no loss of attendance data and no downtime for staff.

I wanted the hosting for a static site defined as code rather than by hand: a private S3 bucket as the origin, CloudFront in front of it with Origin Access Control, TLS via ACM and DNS in Route 53. I wrote the Terraform for every component, Bash and PowerShell deployment scripts, and a GitHub Actions CI pipeline with branch protection on the repository. The stack is authored and in the repository; the AWS deployment is pending and the site is currently served from Vercel.

The core routers at our Hong Kong edge and the core switches in the Philippines carried every client’s traffic over the international link — and both needed replacing. I documented the configurations and dependencies, wrote the cutover with a rehearsed rollback to the old hardware, and failed client traffic over to local ISPs before we touched the core, so clients stayed up regardless of how the change went. We ran it in a 30-minute low-traffic window from a checklist: one engineer executing, one on Zabbix and Cacti, me verifying and acting as the single escalation point for the vendor and the Hong Kong site.

The NOC needed monitoring, DDoS detection, configuration backup, an IPAM/DCIM source of truth and asset tracking — on a budget, after a commercial DDoS subscription had expired. On self-hosted Ubuntu servers I deployed and operated Zabbix for network monitoring, FastNetMon for DDoS detection (replacing the expired Blue Coat subscription), Oxidized for device configuration backup and versioning (LibreNMS as the inventory feed, later replaced by a CSV feed after latency issues), NetBox as the IPAM/DCIM source of truth, and Snipe-IT for asset management with bi-weekly backups, which I also trialled on AWS EC2.

Company policy favoured cloud-first, so the question was whether cloud-hosted monitoring of remote sites over a site-to-site VPN was the right call once running costs were counted. I deployed Zabbix in the cloud and proved it against the Pakistan office over an IPsec VPN, tracked the monthly running cost, and found it uneconomic. I recommended an on-premises server with only SNMP traffic crossing the VPN and delivered the finding as a written cost and recommendation report to the IT Manager.

The department had no formal picture of where its incident, request and change handling fell short of SLA and ITSM expectations. As a solo project I assessed the current processes against those requirements, documented the gaps and produced the action plan. The analysis and plan were delivered — roughly 90% complete at my departure, with the final sign-off meeting pending.

Store managers across the Snoomart dark stores reported slow connectivity and ageing equipment. I gathered requirements with the store managers, audited the assets for replacement or upgrade, verified the reported slowness rather than taking it on faith, and right-sized each store’s internet bandwidth.

I wanted a portfolio that felt distinct from generic templates while staying fast, accessible and easy to maintain. It is a fully data-driven Next.js App Router site with Tailwind v4 design tokens and modular section components: content lives in typed constants, deploys trigger on merge, and updates take minutes instead of hours.