I managed a team at Google, building diagnostics and health management systems for datacenter and ML infrastructure. My focus was on ensuring highly-available and cost-efficient AI/ML clusters for large-scale models like Gemini.

I manage a team building Alibaba's control & management platform for open datacenter-interconnect (DCI) infrastructure. It is a highly-available, distributed system that manages Alibaba's global datacenter connectivity, and provides key features such as resource modeling, real-time monitoring and configuration, inventory management, streaming event processing, business intelligence.Session chair for large-scale infrastructure DevOps at QCon (International software engineering conference)

My work and role remained the same as in LinkedIn, which was acquired by Microsoft in June 2016.

- A member of LinkedIn's data infrastructure team. Working on Ambry - LinkedIn's geographically distributed, highly available and scalable object store serving 500+ million users and billions of objects. Ambry is the source-of-truth store for media and other immutable content at LinkedIn.- Led design and development of key components such as client streaming, end-to-end containerization, etc. - Invited speaker at QCon(International Software Engineering Conference) for Ambry.-

Tech lead, core team member, and key contributor to several projects.1. Performance monitoring for cloud-based network function chaining; - Designed a monitoring as a service (MaaS) architecture for service function chaining; - Architected a virtual probe for dynamic service profiling; - Developed ONF AppFest PoC (best award); - Developed a highly-scalable system for monitoring rule consolidation.2. High-performance cloud and networking - Designed a framework for workload prediction and optimization using machine learning; - Designed a load balancing strategy for multi-stage distributed cloud for latency minimization; - Extended an open-source cloud simulator for streaming traffic analytics; - Built an OpenStack testbed for virtualized service chaining.3. Scalable data centers for service-chaining provisioning - Designed a highly-scalable service chaining architecture using wavelength switching; - Designed and implemented an optimization scheme to reduce service chain provisioning cost; - Designed and implemented an algorithm for service placement optimizing inter-pod communications; - Presented at Open Networking Summit'14, ACM SIGCOMM HOTSDN'14, etc.4. External Communication committee member at the Ericsson Research level. - The only representative from IP & Transport; - Defined and executed strategies for Ericsson Research external communications; - Ericsson Research blog contributor.

Tech lead for the following projects.1. Software-defined networking (SDN) - Designed OpenFlow extension to support elastic optical networking; - Designed and implemented key optimization algorithms for SDN control plane;2. Converged mobile/fixed metro network - Developed optimization algorithms for high availability of converged metro networks; - Key contributor to the design of GUI-based PoC;3. University collaborations. - Supervised 3 intern students.

Worked on metro converged networks. Main contributions include: 1. Designed and implemented algorithms for access network virtualization; 2. Developed a java-based simulation platform for performance evaluation.