Xinlin (Marcia) Yang

PhD Student in Computer Science

Singapore Management University

Brief Bio

I am currently a PhD student specializing in AI in healthcare. I leverage vision–language models, multimodal learning, and explainable AI to build systems that can interpret complex medical data and support clinical decision-making. I am committed to advancing human-centered AI solutions that foster trust, reliability, and impact in clinical practice.

Outside of research, I am an avid enthusiast of swimming🏊🏼‍♀️, badminton🏸, surfing🏄, and skydiving🪂. If you share similar interests or have ideas for collaboration, please feel free to connect with me!

Research Interests

Vision Language Model
Multimodal Learning
Explainable AI
Human Computer Interation

Education

PhD in Computer Science, Current
Singapore Management University (SMU)
Master of Engineering in Machine Learning and Data Analytics, 2024
University of Toronto
HBSc in Statistics, 2022
University of Toronto

Recent News

All news»

[Aug 2025] 🤖 Started PhD studies at SMU

[Mar 2025] 🥳 Received PhD offer with full scholarship from SMU in Computer Science

[Nov 2024] 🎉 Proudly graduated from the University of Toronto with a Master of Engineering in Machine Learning and Data Analytics, achieving top grades

Experience

Research Assistant

Medical Vision and Artificial Intelligence Technologies (MVAIT Lab)

Sep 2024 – Apr 2025 Singapore

Funding Coordinator

Honda Canada Finance Inc

May 2023 – May 2024 Markham, Canada

Research Assistant

University of Connecticut

Feb 2023 – May 2023

Business Analyst - WPE

Royal Bank of Canada

Aug 2022 – Sep 2022 Toronto, Canada

University of Toronto Chinese Student and Scholars Association (UTCSSA)

President

Mar 2021 – Mar 2022

Marketing Director

Mar 2020 – Mar 2021

Marketing Associate

Mar 2019 – Mar 2020

University of Toronto Chinese Student and Scholars Association (UTCSSA)

Toronto, Canada

Projects

Some of my projects, enjoy!

Breed and Service Dog Image Detection

Developed an advanced image recognition system to classify dog breeds and identify service dogs, leveraging ResNet50 for transfer learning and data augmentation techniques. This project delivered a robust, deployable model on Huggingface, designed to support applications in pet adoption, veterinary care, and public safety through efficient and accurate image-based classification.

Apr 2024

Predictive AI Model for Optimizing ETF Portfolio

Enhanced financial modeling techniques by integrating ARIMA-LSTM for systemic risk assessment in U.S. equity markets. This project refined traditional Absorption Ratio calculations, achieving higher precision in asset correlation predictions. The model’s performance was validated through back-testing, resulting in an improved Sharpe ratio, demonstrating its potential for informed portfolio management and risk-adjusted

Dec 2023

Hand Gesture Recognition (ASL)

Designed and implemented a convolutional neural network (CNN) for American Sign Language recognition, using PyTorch and transfer learning from AlexNet to maximize model performance. The project achieved a 92.17% accuracy rate, demonstrating potential applications in assistive technologies for enhanced ASL communication through reliable gesture classification.

Jan 2024 – Feb 2024

Scalable Gender Dynamics Analyzer

Developed a scalable data pipeline in Azure Data Factory for automated analysis of career data, examining gender dynamics across job categories and sectors. Using Azure SQL Database for data processing and in-depth SQL queries, I generated insights into gender disparities and employment trends, supporting data-driven strategic planning for workforce diversity and equality.

Mar 2024

Sentiment Analysis for Movie Reviews

Implemented LSTM and BERT models to capture complex sentiment nuances within the IMDB Movie Review Dataset. Leveraging positional embedding and learning rate scheduling, the models achieved 84.71% accuracy, highlighting their capability to interpret emotional subtleties and sentiment dynamics in large-scale text data, applicable to sentiment analysis tasks across industries.

Apr 2024

AI Navigation for Optimized Pathways

Applied reinforcement learning in the FrozenLakeNotSlippery environment to develop an optimized navigation policy, reducing hazard encounters by 50% while ensuring path efficiency. Utilizing a value iteration algorithm, the model iteratively refined state values to achieve policy convergence, underscoring the project’s advancements in reinforcement learning for safe and reliable AI navigation.

Nov 2023

Working Paper

Canada’s Innovation Strategy Development

Conducted a machine learning analysis on the Global Innovation Index, translating insights from top-performing countries into strategic recommendations aimed at elevating Canada’s innovation rank. By training regression models and evaluating critical feature coefficients, I provided data-backed guidance to support policy-makers and business leaders in fostering economic growth and innovation competitiveness.

Discrete Mathematics in Sudoku: Combinatorics and Graph Theory

Performed a mathematical analysis of Sudoku puzzles using discrete mathematics, focusing on combinatorics and graph theory to quantify possible grid arrangements and solve puzzles with unique solutions. By leveraging permutation, combination, and graph coloring techniques, developed an efficient method to address puzzles with extreme clue scenarios. Demonstrated the applicability of mathematical modeling to real-world problem-solving, providing insights into the logical structure and constraints of Sudoku puzzles.

Service

Workshop Support (Data Analytics)

STEM Streams | Society for Canadian Women in Science and Technology

Mar 2022 – Mar 2023 Toronto, Canada

Peer Mentor

Hart House Mentorship Program

Oct 2022 – Mar 2023 Toronto, Canada