* Discord voice chat over XBox and PlayStation5.* AV1 video compression for Discord video call and screen share* HEVC video codec support on mobile platforms* Screenshare clipping

* Re-architecting Netflix's video processing applications over Cosmos microservice platform* Video encoding for interactive playback

Worked at YouTube Media Algorithm TeamVideo encoding system for HTTP-based adaptive video streaming• Developed a new algorithm for encoding bitrates optimization, implemented the algorithm in YouTube video encoding pipeline• Reduced 10+% streaming bandwidth for YouTube• Google 2017Q2 Golden Perfy AwardVideo quality prediction for User Generated Content• Developed a no reference video quality assessment algorithm for user generated video content (ICIP 2017)• Developed a no reference quality assessment algorithm for video distorted by spatially correlated noise (ACM MM 2016)

• Design and develop the firmware that drives the video processor in Qualcomm Snapdragon chipsets, the most widely used smart phone SoC in the world.• Develop and implement frame-rate scalable video encoders such that the video bitstreams of different frames rates can be generated simultaneously. • Develop various features for Snapdragon HEVC encoders, including dynamic frame rate adaptation, long term reference frame and dynamic intra-coded frame insertion.• Develop and implement bitrate control algorithms for real-time video encoding. The bitrate of the encoded bitstream can be adapted to arbitrary value in real-time.

• Work on Video Aware Wireless Networks (VAWN) project, Co-funded by Intel and Cisco.• Conduct subjective study and investigate how to provide great quality of experience (QoE) to wireless video users.• Propose a dynamic model which can accurately track wireless video users' QoE.• Propose a video bitrate-adaptation algorithm and a user admission control algorithm that optimize wireless video users’ QoE. The proposed scheme can reduce network resource consumption by 40% over conventional bitrate-adaptation algorithms without sacrificing users’ QoE.

• Develop a non-reference visual quality assessment algorithm for wireless videos that suffer from data packet losses.• Using this algorithm, the smart phones can automatically predict the perceptual quality of the received/corrupted video without referring to the original/uncorrupted video at the video server. The smart phone thus can request necessary but not excessive transmission resources (e.g. power or bandwidth) from the service provider to achieve great perceptual video quality for users.• Implement an algorithm which measures the blockiness of encoded video frames without referring to the original video content.