"Many failures are not due to limited capacity, but because they do not adhere to the end"
— No Game No Life
Education
Oregon State University, Corvallis, OR
Master of Science in Statistics
Oregon State University, Corvallis, OR
Master of Science in Mathematics
June 2022Oregon State University, Corvallis, OR
Bachelor of Science in Mathematics
June 2017Online Classes
- Computer Science and Programming Using Python @ MIT
- Computer Science 101 @ Stanford University
- Machine Learning @ Coursera
Projects
Research Experience
Denoise Speech Using Deep Learning Networks and Wavelet
Advisor: Adel Faridani, Ph.D. Department of Mathematics, OSU
- Built Encoder-Decoder neural network models by using Convolutional layers for audios denoising
- Applied wavelet for deconstructing and reconstructing noised audios
- Implemented Mallat's algorithm to construct Daubechies wavelets
Sentiment Polarity Identification in Financial News On-going
- Applied Beautiful-Soup to web-scrap daily financial news from major news websites
- Built a robot to collect Twitter daily top news
- Sentiment Analysis & Potential Statistical Methods
Kaggle Data Science Competition
Avito Demand Prediction
Top 2% (1873 teams)Applied pre-trained CNNs to extract features from image data. Stacking of NN and LGN models, trained stack with XGB, and blended with LGB stacker 1:1. Used TF-IDF and PCA to reduce dimensions from 2000 to 200 for Russian texts.
Toxic Comment Classification
Top 2% (4551 teams)Overcame hardware limits by stacking/blending over 20 Neural Network models. Best single model was an RCNN based on bi-gru-cnn-poolings with custom tuning.
Mercari Price Suggestion
Top 8% (2380 teams)Applied TF-IDF from sklearn to extract features from texts. Built a ridge model combining original numerical features and TF-IDF outputs for ad item price predictions.
Quora Insincere Questions
Top 20% (1401 teams)Kernel-only constraint (13Gb RAM / Tesla 80k). Built an RNN with poolings and attention. Integrated a Capsule layer between GRU layers to optimize training under 2 hours, using Stochastic Weighted Average (SWA).
Web Development
Bitcoin Price Time-Series Study
A pattern analysis using D3 to map line charts from Coindesk's closing pricing. Leveraged a dynamic brush zoom function from the D3 gallery. Handled data switching transitions by cleanly removing prior SVG structures from the DOM before generating new renderings.
Experiences
Graduate Teaching Assistant
Sept. 2019 — PresentOregon State University
- Teaching assigned College level Statistics/Mathematics courses
- Assessing and correcting students’ programming code
- Leading class discussions and answering student questions
- Evaluating students' written answers on assignments
Skills & Stack
Curated Study Resources
- Algorithms
- Introduction to Algorithms by Cormen, Leiserson, Rivest, Stein
- Machine Learning
- Pattern Recognition and Machine Learning by Christopher Bishop
- Deep Learning by Goodfellow, Bengio, Courville
- Web Dev
- HTML and CSS: Design and Build Websites by Jon Duckett
- W3Schools platform