Projects
Optimizing Magnetic Field Trap Computation for Antihydrogen
2023 spring
- Project done in Berkeley Plasma Antihydrogen Physics Research Group (Professor: Jonathan Wurtele)
- Used machine learning to simulate the magnetic field needed to trap antihydrogen that is created by Antihydrogen Laser Physics Apparatus (ALPHA) collaboration at CERN
- Precomputed magnetic field accurately by Biot-Savart model from ALPHA-g experiment, visualized B-field distribution, and fit to multiple machine learning models with data augmentation for comparision
Physics-Informed Neural Networks
12/2022
Used Jax to implement and reproduce the key concepts and frameworks of the paper Physics-Informed Neural Networks.
Baby Cry Detection and Classification
09/2023 - present
- Developed an application that can identify baby cry audios and the reason of baby cry, based on Dunstan Baby Language
- Built a full MLOps pipeline with a blend use of Docker containers, DVC version control, Weights & Biases experiment tracking, Kubeflow, Vertex AI serverless training, google cloud vm and storage, APIs, and frontend implementation
Constrain Cosmological Parameters using Supernovae
04/2023
Used supernovae data to fit equations to constrain cosmological parameters.
Object Recognition
04/2022
Used Pytorch to train a CNN Network on CIFAR-10 with data augmentation, regularization, dropout, and weight initialization; top 10% on Kaggle competition
Pintos Operating System
2021 fall
Built a Pintos operating system that has user processes, multi-thread managements, and file systems in C language.
RookieDB
2021 fall
- Built a bare-bones database in Java which supports executing transactions in series
- Implemented support for B+ tree indices, efficient join algorithms, query optimization, multi granularity locking, and database recovery
Pac Man
04/2021
Made Pac-Man game using python and implemented a series of artificial intelligence techniques to play Pac-Man, such as informed state-space search, probabilistic inference, and reinforcement learning.
Emooc
10/2019
A Cal Hacks 6.0 project that help students and instructors identify difficult parts in online lectures by interpreting facial expressions.