Xinran Dong
- Github:https://007dxr.github.io
- Phone:19800310035 / 17759060644
- Email:dongxinran0805@pku.edu.cn
- Wechat:PKU_DXR
EDUCATION
- Peking University / Period of Study: 2019.09-2023.07
- Department: School of Electronics Engineering and Computer Science
- Major: Intelligence Science and Technology
- GPA:3.60 / 4.0
AWARDS
- The First Prize at the Chinese College Students Computer Design Competition
- The First Prize at the National Olympiad in Informatics in Provinces,2018
- The Bronze Award at the National Olympiad in Informatics,2018
- The First Prize of "Huawei Cup" Innovation Competition of Peking University
- Merit student of Peking University in the academic year of 2020-2021
- Outstanding Volunteers for Beijing 2022 Winter Olympics and Paralympic Games
INTERNSHIP
Google - Software Engineer Intern / 2022.07.04-2022.09.23
- I developed a tool ETL2SQL to convert Generic ETL config to GoogleSQL. Configs in GoogleSQL are much shorter and more understandable than that in ETL. I carried out unit tests to cover all corner cases and verify the correctness of ETL2SQL.
- I designed a new Message PredefinedFilter to lessen the quantity of MetricGenerationConfig. I developed a tool to validate configs with PredefinedFilter performs equally to the original.
PROJECTS
SOFTWARE DEVELOPMENT
- ReadingMap:ReadingMap is a browser plug-in that improves reading experience through behavioral data visualization.ReadingMap provides unique multi-function progress bar which was awarded the First Prize at the Chinese College Students Computer Design Competition.
- Functional Calculator: Functional Calculator is a desktop program basing on the QT framework of C++. The user-defined function and mode-switching function improve the scalability of the calculator.
- Patahub:Patahub is a paper sharing and exchange platform. Patahub adopts the C/S architecture. We use REACT framework to build the front-end, use FastAPI framework to process requests and store data at the MySQL database. I participate in the whole process of product design, development and testing.
- "WALL·E Adventure" Game: Self-developed game WALL·E Adventure using Pygame.
- Plants vs. Zombies:Develop Plants vs. Zombies from scratch using JAVA.
FRONT-END DEVELOPMENT
- Criminal Clue Visualization System: Use D3.js, Echarts.js to establish multimodal data interaction platform. Help the official detect disappearances at VAST Challenge 2014 MC1.
- The Visual Inquiry System of Jinshi in Ming Dynasty: Display the distribution dynamics of the Ming Dynasty Jinshi through the visual inquiry system.
- Stray Cat In PKU: Considering that Stray Cat In PKU association only has mobile app platform, our team developed Stray Cat Association web platform based on Bootstrap.js.
DATA ANALYSIS AND AI ALGORITHM
- Analysing Emotional Fluctuations in Novels: Import Python-UIautomator package to remotely crawl novel texts and reader comments from the mobile app. Use NLP technology to extract potential laws of emotional changes in web texts and draw emotional arcs. This work won the third prize of the 30th "Challenge Cup" Interdisciplinary Student Extracurricular Academic Science and Technology Works Competition of Peking University.
- Train Ticket Detection: I combined deep learning algorithm and image processing technology. I accomplished detection task with 98% accuracy after CNN training.
- Google Research Football: With the prospect of a skillful football AI,our team designed RL models and tried DQN,D3QN,PPO algorithms based on TensorFlow. The football AI finally achieved 80% strike.
- Observatory Data Crawl: I crawled weather reports from the website and injects them to the SQLite database. After that I analyzed data in multiple dimensions using Pyecharts.
PUBLICATION
PAPERS
- Understanding the Impact of Visual Factors on the Experience of Loading Web Apps, Internetware 2022: This work points out the visual factors play an important part on the user experience except for page loading time. We put forward several effective visual factors and give some suggestions to improve web page rendering.
- Breaking the Memory Wall for Resource-Efficient On-Device Machine Learning, Mobisys 2022: This work introduces the Melon law to save memory resource during DNN training. When the memory runs out, we evict the highly cost tensor in greedy strategy. When the evicted tensor is needed, the target tensor is recalculated from the existing tensor in memory.
- Adaptive Compression of 3D Models for Mobile Web Apps, Mobisys 2022: This work proposes an adaptive compression method for 3D models to speed up the loading process of 3D models in Web Apps.