Welcome to my website.
Computer Science Master, Hong Kong, Sep 2019 - Dec 2020.
- Achieved outstanding academic performance (GPA: 3.6/4.0).
- Got high scores in Data mining, Cluster and cloud computing, Deep learning, Image processing and computer vision, UML .etc.
Internet of Things Engineering Bachelor, Beijing, Sep 2014 - Jun 2018.
- Graduated from the Joint Bachelor Degree Program, and awarded a Bachelor’s Degree from BUPT and a Bachelor of Science with Honours - Upper Second Class degree from the Queen Mary University of London.
- Won University Scholarship for three consecutive years (2015-2017).
- First prize awarded by the First Innovative Methods Application Competition of BUPT, and Third prize awarded by Innovation and Entrepreneurship Camp Committee with the municipal project, “Multifunctional Guide Cane” (2017).
- Third prize awarded by Creative Award of BUPT (2018).
Chinese: Native language.
English: Full professional proficiency; Passed CET-4 in the freshman year; Passed CET-6 in Junior year.
Programming Skills: Python, Java, C, Linux server, Linux shell, Pytorch, Caffe, TensorFlow.
Programming Skills: Axure, ProcessOn, Confluence, JIRA, Microsoft Office, Photoshop.
Beijing, Apr 2019 - Aug 2019
- Responsible for the preliminary user survey, user requirements, problem analysis of the annotation client Mona.
- Defined and designed products, including frameworks and interactive processes. Completed high fidelity prototype and product requirements documents to support mona in dealing with various tasks such as various graphics labeling, multi-attribute labeling, and classification tasks on pictures, videos, and 3D Lidar data.
- Coordinated and managed the project, communicated with relevant R&D, and QA in time for implementation and deployment.
- Obsessed with customers, proactively contact the operation team and project sponsors for feedback to continuously adjust the product and improve user experience.
Beijing, Nov 2018 - Mar 2019
- Participated in the "Vehicle Mismatch" project: Reproduce the DLA network on PyTorch and tune it. The final test accuracy and recall both reached 93.42% on the Stanford Car dataset; produced a vehicle dataset with more categories (1715 category) and larger data volume (10G), and reached 92% on the new dataset.
- Participated in the "Content Security" project: After analyzing the characteristics of the dataset collected from DIDICHUXING, designed a suitable image classification network to filter pornographic images, and create a healthy network environment.
- Built a PyTorch C++ API for image classification tasks, and assistant complete the implementation of the image classification service algorithm in the DIDI's internal training platform.
Master Final Project, Mar 2020 - Aug 2020
- Developed a mobile application, which can recognize the lipstick from images, do lip digital makeup, and recommend lipsticks to users to match the demand for beauty guide, especially for lipsticks.
- Used React Native to build cross-platform applications, the popular Flask as the back-end framework, and MongoDB as the database based on the characteristics of data.
- The lipstick recognition and makeup modules are realized through a pre-trained image segmentation model face parsing model to obtain the lip area, then performing color conversion calculation.
- The recommendation module calculates weights based on user behavior and uses collaborative filtering algorithms to obtain rankings.
Cluster and cloud computing project, Otc 2019 – Dec 2019
- Deploy a Spark cluster on eight virtual machines, use Hadoop HDFS for data storage, and use the pre-train SE-ResNet model in BigDL to train the vehicle brand recognition model.
- By observing ganglia and xen to tune related parameters, finally improve efficiency through load balancing, realize full utilization of cluster physical resources during training, and increase training speed by three times.
Data mining project, Otc 2019 – Nov 2019
According to the real-time data of a restaurant’s weather conditions, the random forest algorithm is used to train the model to predict the relationship between its profit and the weather conditions, and the model is adjusted based on the 100-fold cross-validation method to achieve an accuracy of 70%.
Undergraduate Graduation Project, Nov 2017 - Jun 2018
- Designed a segmantic segmentation network, Spatial Aggregation Network, which can aggregate context information of feature maps based on different sub-regions and enhance the boundary information of the objects.
- Suitable for various scenarios and produces high-quality results in pixel-level segmentation tasks, achieve 35.56% mIoU on ADE20K dataset.
- Won an excellent grade of TOP1%.
Innovation and Entrepreneurship Camp, Jun 2016 – May 2017
Participated in the Undergraduate Innovation Experiment Project "Multifunctional Guide Cane". Led a five-member team to the First Prize with the project.
As the Minister of Science and Technology, responsible for publicizing student activities, 2015 – 2016.
Enthusiastic about volunteer activities, successively completed various volunteer services in Global Mobile Internet Conference, Beijing Pingxifu Hongli Primary School, Global Internet Technology Conference and Aviation General Hospital, 2014-2018.
Operate WeChect Official Accounts about photography and life, Sep 2019 - Now.