Graph Neural Network Research

</aside>

<aside> 🇺🇸 Evanston, IL

</aside>

<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/b7035242-ef42-4e75-843c-edec3d957614/icons8-back-96.png" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/b7035242-ef42-4e75-843c-edec3d957614/icons8-back-96.png" width="40px" /> I am excited to continue my research journey at the Modern Artificial General Intelligence and Computer Systems (MAGICS) lab led by Prof. Han Liu. Here, we study Machine Intelligence, Data Science, and Decision Making through massive computational experiments. Specifically, we focus on modern AI's cohesive approach integrating uncertainty and logical structure to model complex real-world phenomena and computer systems' capability to efficiently implement learning algorithms. Previously, I have worked on multimodal graph neural networks for stock prediction:

Conducted literature review on multimodal GNNs using time-series and text data for stock prediction.
Scraped online reviews of companies listed in the S&P 500 stock index for training data curation.
Developed tool for feature engineering of time-series data including using Large Language Model (LLM) API for sentiment analysis.

Currently, I am working on adapting state-of-the-art State Space Models for heterogeneous graphs:

Proposed and implemented novel graph learning architecture for heterogeneous data leveraging state-of-the-art ML algorithms.
Achieved linear time complexity for real-world graph data through hardware-aware algorithmic memory optimization.
Reduced inference time by 4-5x with Nvidia Tesla A100 Series GPU on Linux-based High Performance Computing Cluster.
Co-authored “HeteGraph-Mamba: Heterogeneous Graph Learning via Selective State Space Model”, Pre-print, 2024. </aside>