At Twilio, I designed and built the dataset lookup layer for an in-house SMS rules engine processing 200k+ RPS. It replaced a third-party classifier and saved $2.3M/year in vendor and infrastructure costs.
I also built the 7727 spam complaint ingestion pipeline and MMS image classification infrastructure. The latter showed carrier-side detection was missing 32x more spam images than it caught, making the case for in-house detection.
At Meesho, I architected Spectra, a topology correlation engine for the experimentation platform, and built platform services including centralized RBAC and a high-throughput config distribution engine. I also led the service’s migration from EC2 to Kubernetes at Twilio.
