The brief
Annapurna Express is one of Nepal's most-read news platforms. During breaking news cycles — elections, natural disasters, national events — traffic spikes are sudden and severe. The old infrastructure wasn't built for it. Pages timed out. Load times hit 11 seconds. Readers left.
The brief was simple: fix it. The timeline was not: six weeks before a major national election, which would be the highest traffic event of the year.
What we found
The audit revealed the classic compounding failure: no CDN, unoptimised database queries running on every page load, images uncompressed, no caching layer, and a hosting setup that wasn't horizontally scalable. Any one of these would cause slowdowns. All of them together caused collapses.
We prioritised ruthlessly. Week one: CDN implementation and static asset delivery. Week two: Redis caching layer for high-frequency queries. Week three: image pipeline — WebP conversion, lazy loading, next-gen formats. Week four: database query optimisation. Week five: load testing, stress testing, edge case handling. Week six: election monitoring setup and go-live.
We were running load tests simulating 50,000 concurrent users the week before a national election. That's not a drill.
The result
Election night. Record traffic. Sub-200ms response times throughout. Zero downtime. The platform served 12 million readers across that news cycle without a single reported outage.
The lesson: infrastructure work is invisible when it works. That's the point.