FlashBlade Object & Data Protection.- S3 over RDMA - boost the performance of AI training and inference workload- Object multi-site strong consistency- Industry's first native Bucket Snapshot- Multi-platform Object Storage- File-Object multi-protocol access

Lead two software development teams in two different time zones, with 20+ engineers running multiple projects in parallel.We successfully delivered the following projects in the past few years:- Anti-ransomware storage capabilities- S3 Object Replication feature: replicates data to different geographic locations for protection and disaster recovery purpose- S3 Object Versioning feature: achieves native data protection for object store- S3 Object Lifecycle feature: delivers data manageability at huge scale- IAM User Policy: access control based on user identity- Data resiliency improvement project: makes the system 125% more reliable and recover 31% faster from failures

Engineering:- Build the initial GA version of high-performance on-prem object storage that is 100% compatible with Amazon S3, but 10x faster.- High availability (HA) for distributed transaction system that is resilient to node failures.- Auto-recovery mechanism that can heal the redundancy of data after permanent hardware failures.- Flash health and wear leveling.Field:- Drive beta testing program for new features.- Perform live demo for new product in front of customers.- Present at Keynote of annual company customer event.- Communicate with customer CIO and engineering team to start POC.- Provide technical training to channel partners.

Design and develop the network virtualization system to provide better flexibility and configurability for datacenter network. Experience here has equipped me with strong programming skill, solid knowledge in networking protocols and the ability to develop, debug and maintain distributed software systems.- Elastic Virtual SwitchAs one of the initial key members of this product, I designed and implemented every building block from scratch along with 2 teammates. Elastic Virtual Switch (EVS) provides the datacenter admin a faster and easier way to deploy and manage the public/private cloud. EVS can manage and monitor L2/L3 network resource among millions of virtual machines across the datacenter as if they are all connected to one intelligent "elastic" switch. EVS can be managed through both CLI and OpenStack. Enterprises can deploy their virtual datacenter network with EVS to significantly reduce the time and cost spent in network resource provisioning, detect connectivity issue easily and recover network configuration from failure automatically. The first version of EVS was released as one of the highlighted key features in Solaris 11.2.

Designed and implemented a NS-2 based Wireless Scalable Stream Video Emulation Framework, which takes video as input, emulates real wireless environment and generates video as output. This project requires knowledge in scalable video coding, NS-2, MAC and physical layer properties and understanding in protocols such as TCP/IP, UDP and AODV. This project was supervised by Dr. Ritesh Madan from Qualcomm and the work was very well received.

Conduct research in the Wireless and Mobile Networking Lab of NTU. We designed and implemented a special wireless network over public transportation system. We use accelerometer to detect the behavior of vehicle and design our algorithm to improve transmission rate. The result is published as a poster in Sigcomm 2010.