Hello there ! I am

About Me

4 years building B2B service platform for over 100k+ users.

Expert at turning complex business logic into manageable and modularized components.

Adaptable to new technologies and unprecedented feature requests.

Strong and concise communication skills to gather information and convey thoughts.

Tech Stack

Frontend Frameworks

Backend Frameworks

UI & Styling

DevOps & Deployment

Database & API schema

Data Processing

Data Streaming

Online Tools

Testing

Languages

AI

Others

Experience

Education

New York University (Graduate Meng)

Computer Engineering

To expand interests and expertise from my previous work experience, I focus on broader study of embedded systems, robotics, and deep learning, tapping into the possiblities of post-AI applications and trends.

2025 / 09 ~ current

National Taiwan University (Undergrad)

Chemistry

Focused on the study of quantum chemistry and computational chemistry. Completed my personal project on 2d spectrum analysis with Convolutional Neural Network (CNN). Received a grant from the Ministry of Science and Technology for my research.

2016 / 09 ~ 2020 / 06

Taipei Municipal Chien Kuo High School

Explored the overall development and interest in all fields of science, including chemistry, physics, mathematics, and computer science.

2013 / 09 ~ 2016 / 06

Have you ever regretted a message after sending it?

Portfolio

Tech Talks

Model Context Protocol

MCP clients, servers, transport mechanisms, and use cases

Zero-Friction Page Experience

with Speculation Rules API + View Transitions API

React Fiber : A new way to think about React

What's new in React 18 ?

Brief introduction to trpc

Think like google search : How to search efficiently ?

Brief Intro of Astro Framework

Research Projects

2026/04/05

Location: New York University (NYU)

↗

Prompting and Fine-Tuning Qwen for Text-to-SVG Generation under Structured Output Constraints Authors/Creators

We study text-to-SVG generation as a structured sequence modeling problem in which a model must map natural language descriptions to valid, compact SVG markup. Our approach combines dataset analysis, SVG preprocessing, prompt design, and iterative model ablations under practical hardware constraints. We evaluate multiple configurations across AMD MI100 and Kaggle NVIDIA T4 environments, spanning Qwen2.5 instruction-tuned and coder-specialized model variants, different context lengths, LoRA settings, and training data scales. The strongest configuration uses Qwen2.5-Coder-3B-Instruct with a 3072-token context window and achieves 10.77991 on the private leaderboard and 13.11993 on the public leaderboard. Our results suggest that preprocessing quality, task-aligned prompting, and selecting a model family suited to code-like generation were more important than simply scaling baseline hyperparameters in isolation

2026/03/14

Location: National Taiwan University (NTU)

↗

Classification and Interpretation of Two-Dimensional Electronic Spectra Using Convolutional Neural Networks

This study applies a convolutional neural network (CNN) to the interpretation of two-dimensional electronic spectra (2DES), and establishes a program framework covering spectral-data preprocessing, model training, and testing. Using simulated 2DES as model input, this work interprets excitation-state energy difference (∆), coupling constant (J), and dipole angle.