• Tech lead for Kafka 8

• Worked on storage layer for Cherami (Uber's durable and scalable task queue engine)

• Tech lead for personalized predictions in Uber Maps (powers 55~60% of all trips for pickup and destinations)

• Tech lead for large scale immutable key value store (used by Maps Routing Engine and Tile Display, Uber AI Personalized Predictions platform, etc.)

• Tech lead for 30 person P0 effort to provide real time insights and diagnostics into individual trip data

• Tech lead for infra layer on Project Next in UberAI, Uber's next generation personalized predictions platform

• Worked on the Infrastructure Team, building backend distributed systems for serving data at scale

• Designed and built datacenter simulation running on a single commodity hardware by emulating entire datacenter worth of Hyper-V machines to be managed by the product

• Led in-depth analysis and prototyping effort and improved product-wide performance by many folds by alleviating threading and lock contentions, greedy memory allocations, deficient database schema and heavy network traffic

Hyper-V Hypervisor

• Prototyped and implemented early launch of the hypervisor, focusing on bootstrapping multi-processor systems to initialize physical processors and configure real/protected/long mode memory

• Help drove the performance/scale effort to improve from 24 to 64 logical processors in 6 months

• Analyzed/improved performance in shadow page tables, interrupts, spin locks and NUMA hardware

• Implemented numerous debugger extensions to understand the hypervisor performance