Rain Xu
Somerset, New Jersey 08873 | (732) 485-3714 |
Email:https://sddyxrf@gmail.com | LinkedIn: https://www.linkedin.com/in/runfeng-xu | GitHub: https://github.com/runfengxu
EDUCATION
Rutgers University, New Brunswick, NJ 01/2019 - 12/2020 Master of Science in Computer Science GPA: 3.7/4.0
Fudan University, Shanghai, China 09/2014 - 07/2018 Bachelor of Science in Physics
TECHNICAL SKILLS
Programming Languages: Python, C, Html/Javascript, SQL, CUDA Concepts: Database, Deep Learning, Operating System Programming, Unix, Full-stack
EXPERIENCE
Software Engineer Inter, Vouched, Seattle 05/2020 – now
• Designed and Implemented efficient detector model and landmarking model for edge-devices, the final model is only 100Kb
and able to perform real time detection on low end mobile devices. Wrote testing web application and benchmarked speed
and resource consumption in TensorflowJS+NodeJs developing environment.
• Updated the current text recognition model by combining semi self-learning based model with EfficientNet Structure,
improved the accuracy on our customed dataset from 70% to 95%.
• Deployed a online cross-platform multiple poker player written in redis, node.js and socket.io as the social event for
company’s chat time.
Research Assistant, Rutgers Computational Biomedicine Imaging and Modeling Center, Piscataway, 08/2019 – 11/2019
Non-manual features recognition in American-Sign Language,
• Used CNN+LSTM based neural network to extract non-manual features( including eyebrow motion and aperture motion) from 1000+ sign videos
Manual Features detection in American-Sign Language
• Used proposal-based neural network to detect and segment manual-sign event in temporal scale, eventually recognize all manual sign events and their meaning in over 1000 sign videos.
Data Engineer Intern, Minsheng Bank, Shanghai, China 07/2018 - 08/2018
• Crawled data from real estate agent websites and e-commece platforms through Python Scrapy framework. Used crawled data to analysis regional consumption level to help the bank provide personalized services for their customers. • Build a visualization of the analysis using Baidu map API.
Intern, China Union Pay, Shanghai, China 07/2017 – 08/2017
• Extracted keywords from documents in database using TFIDF algorithm and built a documents matching algorithm to search for similar document in the database. With similar old documents for referencing purpose, my colleagues could handle new requested query more efficiently. • Build a web frontend for searching matched documents using ASP.net and SQL Server.
PROJECTS
NCAA online predicting engine, Rutgers (AWS EC2/S3, Mysql, Redshift, mongoDB) 04/2019
• Collected NCAA tournaments and final games stats for the last 10 years. Designed and implemented a database query engine using redshift and MongoDB, We build a game prediction model using KNN algorithm over the data and a webpage to support online stats searching and predicting. the service is built on AWS cloud service.
Colorization Black-White photos, Rutgers (python) 06/2019
• Designed a neural network based on CNN to colorize gray photos, all code including forward/backward propagation are hand-written.
Rail Track Detection, Rutgers 05/2019
• Used python to detect rail track in the video and use Yolo framework to segment the pedestrian and vehicle in the same view
CUDA PIFO task scheduler, Rutgers 12/2019
• Designed and implemented a GPU scheduler on CUDA runtime level which combined the principle of PIFO scheduling policy and CUDA asynchronous executing fashion. It decides the order in which received tasks are executed to use the limited data transfer pipeline and GPU kernel resource more efficiently and reduce overall run time and response time.
E-commerce recommending system, Rutgers 05/2019
• Designed and implemented an online shopping and recommending system using python an flask. The data were from a Brazil shopping platform which includes over 1000 items and 70k transactions. The recommending engine is built using AR model and Apriori Algorithm.
