01 · Inherited a platform under pressure
This wasn't a redesign. When I joined in October 2023, FMT was a live newsroom losing ground. An earlier migration had left the platform brittle, and the numbers showed it: traffic and visibility were sliding, pages were slow, and sections of the site went missing exactly when traffic spiked.
The hard part of inheriting a system like this is that the symptoms are business symptoms — declining reach, weaker ad performance, frustrated editors — but the causes are technical and tangled together. My job was to own the platform layer and make the whole stack work as one system again, without ever taking the newsroom offline.
02 · What was actually broken
The problem was never one bug. The earlier migration had shipped without the SEO foundations a news platform lives on — structured data and JSON-LD, proper metadata, clean URL behavior, and Core Web Vitals that were nowhere near green.
Underneath that, the CMS was being throttled by the database: slow queries, missing indexes, connection limits and memory/CPU pressure that pushed load consistently around 50% and often to 100% during spikes. When the database choked, page sections simply didn't render. On a breaking-news day, that's the worst possible failure mode.
03 · What I changed
I set the Next.js platform direction and rebuilt the rendering and caching model around it — ISR for news content, with custom caching designed to work with Cloudflare rather than fight it. I repaired the SEO layer from the ground up: metadata, structured data, schema and URL behavior, so the platform could be indexed and surfaced properly again.
On the data side, I refactored queries, re-indexed, and added a custom cache compatible with both Next.js and Cloudflare. On the edge, I hardened security with Cloudflare firewall upgrades, bot management, audits and patches. And throughout, I pushed FMT toward independence — building in-house tools and migrating services to self-managed platforms to reduce external dependency.
04 · What changed (results)
The platform moved from a declining narrative to a growth one: supporting 8.5M+ users at scale, with 184% organic growth in the public turnaround story and load time improved from 5.2s to 1.3s. Database load came down from the 50–100% danger zone to around 30%, with only rare spikes toward 40%.
Visibility recovered alongside it — the platform reached Google News publisher status and subscriber channels grew off a near-zero base. The cloud bill came down from RM45–60K a month to RM12–15K — RM400K+ a year back in the budget — by making the platform talk to the database less and the cache more.
05 · Why it matters
This is the clearest evidence I have of hands-on technology leadership: diagnosis, architecture, engineering, database cleanup, edge strategy, security hardening, vendor reduction — and enough rollout judgment to keep a live business running through all of it.
And it's not just a media story. Any operator — a publisher, a marketplace, a bank, a logistics platform — recognizes the shape: you inherit a system under pressure, find the technical causes behind the business symptoms, make SEO, cache, database, frontend, CDN and security work together, reduce dependency, and keep the business running the whole time.