• Implemented smart cropping algorithm on shopping blocks through c++ in back end interacted by +800K users• Replaced original images with region of interest on shopping blocks in java front end to improve quality• Launched and analyzed smart cropping and brand logo experiments on 1M users• Investigated and implemented algorithms to select more diverse and representative images o

Course Info: Logic Circuits, Linear Algebra• Guided students to discover solutions to problems by preparing additional materials and answering related questions• Taught the fundamental concept of Verilog for over 50 students to design circuits

• Consolidated related works of lifelong learning including memory replay method,regularization approaches and changing architectural properties to 10 team members• Implemented lifelong method A-GEM model through Pytorch to solve unlabeled data problem and provided thorough evaluations in different specs of data such as Split CIFAR

Publication:An Experimental Evaluation of Recent Face Recognition Losses for Deepfake Detection 2020. ICPR• Achieved 95+% accuracy on forgery detection of millions of human faces data with CNN based and LSTM network applying Arcface loss, L2softmax loss, and focal loss• Provided a thorough quantitative performance evaluation, including temporal aggregation method, score-level fusions, and t-SNE feature representative visualization

• Developed the Web Interface in Vue and maintained the stability for NAS users• Integrated and summarized technical documents reached 100+ research developers