From 200 to 5,000 learners in three months

snappyloop is an AI language-learning product — courses, live voice practice with a tutor, and spaced-repetition vocabulary, delivered through Telegram and its Mini App.

I built the platform: the infrastructure and security for the multi-agent system behind it, load-tested at 20,000 requests per second. As the product grew from 200 to 5,000 users in three months, the architecture moved from a single long-polling worker to webhook-driven, horizontally scaled bot workers — with Redis for shared session state and global rate limits, and a central controller holding one ceiling across every replica for the AI vendors.

Security is four layers deep on the webhook path — firewall, ingress, IP allowlist, and request-signature verification — with the internal API reachable only through an outbound-only Cloudflare Tunnel. SLO burn-rate alerts on webhook acknowledgement latency and reply success rate, plus multi-city Playwright probes, watch it in production.