This article was written by our CTO & Co-founder, Michael Monin.

<aside> ℹ️

Why this matters for candidates

We're not a prototype. We run one of the most active AI companionship platforms in the world, on a lean, fast-moving Engineering team. If you care about scale, ownership, and shipping, read on.

</aside>

Scaling AI companionship — under the hood of a 10B request monolith

💎 Why We Use Rails

Garry Tan recently noted that the world is "sleeping" on the combination of Ruby on Rails and AI-augmented coding. Because Rails is built on "Convention Over Configuration," AI models like Claude can act autonomously within the codebase, resulting in massive productivity gains.

At EverAI, we’ve turned that "unlock" into a reality at massive scale. We aren't just building an app; we are running a global infrastructure that handles:

• 28,000,000 Monthly Active Users.
• 10,000,000,000+ HTTP requests every 30 days.
• 1 Petabyte of monthly data traffic

📊 The raw numbers

10–12 billion HTTP requests in a typical 30-day window. Peaks push well north of 400M/day. 80M+ visits, 41M+ uniques, and we're serving a full petabyte of data (mostly images/videos — they're heavy).

Last 30 days - Requests

⚒️ Keeping the queues happy

Everything non-trivial is async: image, messages, notifications, ML inference queues, analytics events etc.. All flowing through Sidekiq.