The challenge
Traffic would arrive in bursts driven by a national oral health campaign. A two-person engineering team needed to keep the working platform available while finding the bottlenecks behind slow responses.
My work at Logy.AI connected backend engineering with an AI-powered screening experience delivered through WhatsApp. The campaign needed to handle sudden demand while continuing to support the people already using it.
The decisions
Improve the database lookup path, reuse connections, cache repeated reads, and move reminders and analytics into independent event consumers. Each change addressed a specific source of load.
The request path stayed focused on saving the screening and returning a response. Event consumers handled reminders, analytics, and conversion tracking separately. That made it possible to add or retry background work without making every user wait for it.
Keep the user response path clear.
The outcome
The India campaign reached 4.5 million people. My published engineering account describes 550+ requests per second sustained, while the database was reduced to one-eighth of its earlier size.
These are historical results from the campaign described in the article. The broader healthcare platform and the India campaign have different scopes; the figures here refer to the India campaign.
The lasting lesson was to understand the bottleneck before buying more capacity. A targeted index, connection pooling, caching, and a separate execution path for background work changed how we could operate the system.