Bellevue, Washington, United States
Building hyperscale AI, cloud, and distributed data platforms from 0→1 and scaling them globally. I lead Microsoft CoreAI Inferencing, a multi-geo engineering organization that powers flagship Microsoft products—including GitHub Copilot, Microsoft Copilot and LinkedIn—along with some of the world’s most innovative startups, unicorns, and thousands of enterprise customers.
• Lead the core inferencing team building the largest SaaS inference service and scaling it through exponential growth. • Drove the vision, strategy, and execution of inferencing as a SaaS offering, collaborating with internal and external stakeholders, product managers, and architects. • Partnered with product team to develop new customer offerings (Global, Data zone and regional consumption and provisioned offers more than doubling the YoY revenue) • Achieved 30% platform inference efficiency, contributing tens of millions in margin improvement highlighted in MSFT FY25 Q1 earnings through novel routing, caching and scaling (7 patents filed). • The service supports inferencing all Microsoft external customers including top AI startups as well as all Copilots (GitHub Copilot, Microsoft Copilot, M365 Copilot, Security and LinkedIn)
• Architected and scaled Azure Cosmos DB service 100x. • Delivered compute platform for Cosmos DB service for compute-intensive, dedicated and multi-tenant workloads like gateways, caches, Mongo and Cassandra translation engines. • Developed Cosmos DB Database accelerator for ultra-low latency, cost-optimized data access. • Increased adoption of Azure Cosmos DB in F500 segment by building enterprise-grade networking and security features (PrivateLink, RBAC, Network Isolation). • Ensured customer success on Azure Cosmos DB.
• Built core Messaging and Calling services for mobile collaboration • Service features: Sync, Push notifications, Slack integration, Web auth, Call export, User status, Service support for materialized views. • Post acquisition by Microsoft, Principal SDE (Dec 2015 - Dec 2016), built and launched Microsoft Teams
Built Azure Load Balancer from ground up Designed and created Software Load Balancer (SLB) for Azure from inception to key platform enabling Azure growth. * SLB is a fully distributed system running on all nodes in Windows Azure cloud. Multiple instances of SLBs are deployed in Azure and had a combined capacity of multiple Pentabytes per second. (Description in ACM SIGCOMM 2013 paper: Ananta: Cloud Scale Load Balancing) * Invented and implemented novel mechanism to reduce data path latency by more than 50% and increase throughput by 100x for intra-region Azure traffic. (Details in section 3.2.4 of above paper.) * Scaled SLB to 50 TBps
Built kernel/low level networking/transport features.
Software Developer in early-stage VoIP companies.