Experience

“What I cannot create, I do not understand” - Richard Feynman

Here are some hobby projects that I built either for hackathons, contract project with companies, or simply for fun!



Neighboring state-based RL Exploration

A model-agnostic, on-policy exploration method in deep reinforcement learning. Our method consistently outperforms Double DQN baseline by 49% in test time under a discrete-reward environment. Built in PyTorch, MuJoCo. Read more on ArXiv and GitHub.


Sentiment Analysis for Headline Trading

Dashboard for efficient inspection of market sentiments toward individual stocks based on sentiment classification results on auto-scraped news headlines, using fine-tuned BERT language model. Built in PyTorch and React. Read more on GitHub.


ClassCaster

HackMIT-winning project (Best ML Hack, 2021). A ML-powered workload forecaster providing students with predictions on estimated time for each task and study group recommendations using K-means clustering. Built with React, Node.js and Python. Read more on GitHub.


Cypher AI

CalHacks-winning project (Winner of IoT Security Challenge, 2021). An encryption solution for ML systems on IoT devices, utilizing Fully Homomorphic Encryption for servers to run ConvNets for data-sensitive applications such as face recognition, without decryption. Built with Flutter, PyTorch, OpenCV. Read more on Devpost and GitHub.


TopScore CMS

A Content Management System (CMS) for an educational consultancy firm to manage student database. This CMS transformed company’s doc management from entirely manual work to 100% digital process. This is a contract project with TopScore.


Poetic Dream

Deep Learning web-based project that transforms poetry text into artistic portraits, utilizing techniques and structures such as OCR, Neural Style Transfer, LSTM. Built with Bootstrap, Python, and TensorFlow. Read more on GitHub.


Gitlet

Implemented my own version-control langauge in Java. Gitlet's 20 commands, e.g. `commit`, `merge` and `checkout`, cover most of the essential local and remote features in actual Git. Read more on GitHub.