GPU and TPU supercomputers

https://cloud.google.com/blog/products/compute/introducing-a3-supercomputers-with-nvidia-h100-gpus

Previously, Technical Lead Manager in Google Brain, working on the next generation TensorFlow infrastructure

• Building Uber's globally distributed transactional datastore. This was Uber's home-grown Spanner like db using open source technology like RAFT consensus and RocksDb

• Big data and storage infrastructure for self driving cars

Self driving cars rely on large scale ML training based on log data, which tend to be an extremely large number of very small sized files. Traditional storage systems like PG SQL or on-prem tech like parallel file systems have notoriously problematic scaling and perf challenges. I helped build a cloud-native parallel file systems that was built on commodity object storage targeted for this use case. This kind of tech has since then become a staple in the cloud provider itself (see AWS EFS offering)

Microsoft Azure infrastructure - SDN and NFV for Windows Azure. Worked on various distributed systems infrastructure projects. I was one of the original founders and technical lead for SLB - Azure's distributed layer 4 software load balancer as a Service (SIGCOMM paper https://conferences.sigcomm.org/sigcomm/2013/papers/sigcomm/p207.pdf)

I also worked on several incubation efforts like Azure storage RTable

Distributed Transactions and Coordination protocols, Windows Workflow foundation, Windows Communication foundation

Software Developement - .NET Framework (Visual Studio, Workflow, Transactions), Windows Server 2008 (longhorn server) - Microsoft Distributed Transaction Coordinator (a windows NT sevice)