M.S. Statistics student at the University of Washington and B.S. Mathematics student at the University of Illinois Urbana-Champaign
Here is my playground for sharing some information for people who want to know more about me academically. In my childhood, I always imagine how world will be different when AI and robotic become wildly used in human world. Now, the AI epoch is coming, I am passiobn to see that happen and want to be part of it. Lets work together to make the history 😆
Project-first highlights for fast recruiter scanning.
Problem: Improve segmentation quality for BraTS 2021 MRI volumes under multi-modal uncertainty.
Method: Compared baseline FNO, diffusion U-shape, and an FNO-Diffusion hybrid with aligned evaluation.
Result: Built a case-study workflow showing where hybrid modeling improves robustness across regions.
Problem: Compare RNN baseline quality against pretrained transformers on IMDb sentiment.
Method: BiLSTM vs DistilGPT-2/XLNet with built-in vs MLP heads and max-length ablations.
Result: XLNet reached 0.9578 TestAccuracy; longer context helped with diminishing returns after ~1024 tokens.
Problem: Simulate realistic NPC action choices under hard room-access constraints.
Method: Prompt and task-design experiments with Mistral 7B and Mixtral 8x7B, plus GPT/Claude comparison checks.
Result: Produced a prototype workflow that reduced invalid action outputs on constrained assignments.
Unsupervised learning, representation learning, GLM/LMM/GEE, statistical inference, experimental design, and evaluation metrics.
Python, R, PyTorch, HuggingFace, Shiny, Git, SQL.
Reproducibility, clean code, and clear communication with research and product teams.
Open to applied scientist / ML engineer opportunities and research collaborations.