At Plivo, I built and now own a Python voice-AI agent service handling thousands of production calls a day at 99.9% uptime. I migrated the live voice pipeline across seven services with zero downtime, moving customers one organization at a time behind a feature flag.
I also built a run-tracking subsystem spanning six services to record lifecycle state and per-component cost across voice, chat, and messaging conversations. It supports both customer-facing run listings and the billing pipeline, with concurrent updates designed to avoid duplicates and lost writes.
On Plivo’s Messaging Platform, I developed customer-traffic fraud detection and addressed a platform-wide cache failure risk with cache-aside reads and regional failover, helping messaging services maintain 99.99% uptime. I also built WhatsApp conversation-based billing with idempotent event handling.
Earlier at Plivo, I tuned SMS routing, led an India DND integration, and improved message-processing test coverage from roughly 6% to 60%. My projects include a thread-safe Go in-memory key-value store and scheduler extensions for MIT’s xv6 teaching OS.
