I am Lucas Chan, and I am currently pursuing an MPhil degree at HKUST, where my research focuses on multimodal learning, knowledge graphs, and large vision language models (LVLMs). Under the supervision of Professor Yangqiu Song, I am exploring the intersections of these cutting-edge fields in artificial intelligence. Prior to my MPhil study, I obtained my Bachelor's degree with a double major in Computer Science and Electronic Engineering at HKUST, and graduated with First Class Honours.

Currently, I am expanding my research into multimodal learning, with a particular focus on developing high-quality question generation systems using Multimodal LVLMs agents. This emerging area of study represents an exciting frontier in AI research, and I welcome collaborative opportunities and intellectual discourse on this topic.

My academic work spans several areas of machine learning and its applications. I am the lead developer of Neural Stock Prophet, a PyPI package that integrates advanced algorithms such as LSTM neural networks with attention and Autoregressive Integrated Moving Average (ARIMA) models for sophisticated stock forecasting. The project is currently in its testing and enhancement phase.

My portfolio also includes significant contributions to other machine learning domains. Notable among these are:

• Car License Recognition: An optical character recognition (OCR) system employing transfer learning techniques for accurate vehicle license plate identification.
• Federated Edge AI for 6G: A novel approach combining Federated Dynamic Regularization (FedDyn) and Reconfigurable Intelligence Surface assisted Federated Learning (RIS-FL) to address computation and communication heterogeneity in federated learning environments.

In addition to my core research areas, I am particularly passionate about applying machine learning techniques to quantitative research. This interest stems from my belief in the transformative potential of AI in  financial modeling, risk assessment, and market analysis. My work on Neural Stock Prophet is just one manifestation of this passion. I am actively exploring ways to leverage advanced machine learning algorithms, including deep learning and reinforcement learning, to enhance traditional quantitative models. My goal is to develop more robust, adaptive, and accurate predictive systems that can navigate the complexities of modern financial markets.

• Multimodal learning
• Large vision language models
• Portfolio optimization

• Coding (especially leetcode everyday)
• Tracking the lastest news in AI
• Stock trading
• Playing some video games (currently monster hunter rise from switch and I know it's quite old)
• Gymnasium