Worked on Cabot, which is an assistive robot to help visually impaired users navigate through a new environment with BLE beacon, lidar and vision.

• Designed and built the whole system architecture and pipeline with ROS, including pedestrian detection with Kinect One, navigation stack and other human-robot-interaction components

Worked on the EUREKA: Enhancing Understanding of Real Environments using Knowledge from crowd Assistance system, utilizing crowd power to tag point clouds got from robots.

• Proposed and implemented smart selection filter using Kernel Density Estimation that eliminates accidentally added points, improving average crowd selection time by more than 20%.

• Proposed and implemented smart selection filler that predicts potentially wanted points using bushfire algorithm and k-dimension tree, which is 2000% faster than the previous method.

• Built the UX screen for creating content strategies of Amazon Video with AngularJS 1.x, which is the main entry for all the merchandizers to edit and view the strategies.

• Created a visualize collapsible tree from scratch for metadata in a database using D3.js. The tree is used by developers in Amazon internally to help debug the data in the database.