At CoreStack, I led architecture and proof-of-concept work to replace MongoDB with a sharded ClickHouse warehouse for multi-tenant cost analytics. The new approach scaled processing capability by 50% and supported more than 1 billion rows per run.
At Yotascale, I built FinOps and cost-observability products used by enterprise companies including Hulu/Disney, Intuit, Roblox, Compass, Jefferies, ClickUp, and Okta. I optimized Spark processing to handle more than 2 billion rows, cutting runtime by 30% and cost by 15%.
I’ve also developed forecasting and anomaly-detection solutions for cloud costs, including an anomaly detection spike that helped Zoom save $600,000 in runaway costs. I write about Apache Spark, ClickHouse, dbt, and production-grade MCP servers.

