Hi my name is Kaiwen!
I'm a graduate student at Georgia Tech specializing in machine learning. My undergraduate concentrations were in modeling and simulations, and systems and architecture. In addition, I was part of the Honors Program and International Plan, and received a Certificate in Finance and an ACTFL language certification in Mandarin. I do a mix of full-stack web development along with high-performance data analysis and processing algorithms, with some Android development sprinkled in. My favorite challenge is trying to make sense out of data, either through algorithms that try to predict or visualizations that give the most bang for buck. Outside of computer science, I design parts in CAD and fabricate through machining or through 3D printing. My hobbies include Nerf modding, painting, sim racing, and playing the piano and electric bass.
Quick Stats
Familiar Technologies
Languages
- Java
- C
- Python
- HTML
- LaTeX
- MATLab
- PHP
- MySQL
- MongoDB
- Hadoop
- Spark
- VHDL
Programs
- Microsoft Office
- OpenRefine
- AWS EC2
- Microsoft Azure
- Tableau
- Gephi
- Git
- Linux
- Anaconda
- Vi/Vim
- Android Studio
- Visual Studio
- Autodesk Inventor
- Adobe Photoshop
Frameworks/Libraries
- Django
- Pandas
- Scikit-Learn
- NumPy/SciPy
- Plot.ly
- D3
- Bootstrap
- Zurb Foundation
- MPI
- Nginx
Work Experience
Software Engineering Intern - Hughes Network Systems (Germantown, MD)
- See Sandvine Project for more details
- See Video QoE Project for more details
Research Assistant - Resources Global Professionals (Shanghai, China)
- Aided in fulfilling RFPs for several Fortune 500 companies and various consulting services
- Gained valuabe insight in business functions and compliance regulations
IRTA Fellow - National Institutes of Health (Bethesda, MD)
- Worked in the National Library of Medicine to build a request management system for NLM servers
- See QueueManager Project for more details
Selected Projects
Scene Recognition with Bag of Words
- Implement scene recognition through several methods, including tiny images, nearest neighbor classification, bags of features, and support vector machines
- Dataset consists of 1500 samples from 15 scenes from the SUN dataset
- Able to reach 65% accuracy with SVM, with a vocab size of 400
Fundamental Matrix Estimation with RANSAC
- Estimate fundamental matrix to match SIFT features with greater accuracy by calculating projection matrix and camera center
- Implemented SIFT with a Harris Corner Detector and Adaptive Non-Maximal Suppression
- Successfully matched features in an image with >98% accuracy
Video QoE Experiment
- Survey of video QoE metrics to predict subjective QoE given objective data
- Extended Android app and PHP application to survey users on subjective QoE while recording video network statistics
- Trained and tuned classification algorithms to predict subjective QoE in Azure ML Studio and Sklearn
Box Office Revenue Predictor
- Predict movie revenue based on a number of factors, including past movies, genre, actors, and critic rating
- Used machine learning to build models off multiple data sources
- Created intuitive visualizations to display data and predictions
Audio Transcription
- Python notebook to try and recognize musical notes based off audio files
- Used signal processing to recognize monotonic notes with reasonable accuracy, and to some degree simple polyphony
- Entry for the 2018 Get A Move On Hackathon
Sandvine Dashboard
- Full-stack dashboard to monitor aggregate web traffic data through network gateways, and monitor for anomalies
- Consisted of data retrieval, cleaning, and analysis, followed by visualizations and cluster analysis
- Designed for robustness, flexibility, and speed, due to multivariate and very large data sets, subject to change often in the future
Window Seat
- Virtual Reality application designed for all smartphones to be used on flights as alternative entertainment
- Takes flight data and projects a virtual window seat, allowing interactivity with points of interest outside the plane
- Entry for HackGT 2018 Delta Airlines Hack
ActiveLearning
- Cross-platform web application that allows instructors to pose questions on a board and allow students to answer in real-time
- Implemented review and quiz features that can be enabled by a professor for students to use
- Worked with client to deliver and test-deploy application
Shopping With Friends
- Android application that allows users to add and share deals and discounts with each other
- Used geolocation to enhance user experience
- Designed using full UML development tools in an Agile environment
LC-3220 Processor
- 5-stage pipelined processor design based off a given ISA for an Altera FPGA
- Implemented speedup features such as data forwarding and two-bit branch prediction
- Written in VHDL using Quartus software
HKeInvest
- ASP.NET application built for a mock investment company
- Allowed users to apply for and exchange on a virtual stock market
QueueManager Dashboard
- Full stack web control panel that controls all incoming requests to major library servers
- Built from scratch to replace older panel, using updated technologies such as AJAX and JQuery
Contact
Email: kwang329@gatech.edu