Responsibilities include: Machine Learning for Product Modeling

• Currently working on Uncertainty-aware Deep Learning with SNGP

Responsibilities include: Object Detection Model Building

• Implemented COCO segmentation format and used data augmentation techniques for preprocessing.
• Utilized object detection model YOLO to identify LED screens to enhance clients’ user selection experience.
• Assisted Hon Hai Precision Industry in automation by using Detectron2 model, achieving over 90% accuracy in checking for defects in hardware components like ribbon cables, heat sinks, coaxial cables, and small wiring harnesses.

Responsibilities include: Backend Development and Data Pipeline Optimization

• Developed and maintained integrations with external eSports data providers to verify odds values.
• Engineered and managed pipeline using AWS Lambda, S3, and DynamoDB.
• Designed and optimized the games’ data pipeline to ensure data integrity and usability for predictive analytics.
• Refactored code to remove unnecessary blocks, improving overall code quality and performance.

Responsibilities include: Building a Multi-Agent System for Real-Time Conversations Between Humans and AIs

• Contributed to the development of a Multi-Agent System that facilitates real-time interaction between humans and AI from scratch, enhancing user experience and system responsiveness.
• Optimized the Recommender Systems with 10000 virtual user profile.

Responsibilities include: LLMs for ontology matching

• Applied LLMs such as GPT-4, BERT, etc. to analyze and interpret metamodels; Benchmarked and tested ontology matching algo COMA 3.0 and LogMap on a set of metadata.
• Achieved an average F-measure of 91%, greatly surpassing the baseline approaches, which have a maximum average of 29%

Responsibilities include: Machine Learning for Code and Documentation

• Integrated Large Language Models (LLMs) to improve Software Engineering practices, particularly in code generation and documentation.
• Developed and deployed web-based tools using React.js, facilitating the use of LLMs by software engineers.

Responsibilities include: Software and Hardware Development

• Engineered a 3D-aware localization system that significantly improved spatial accuracy using UWB technology.
• Programmed and refined the collision detection algorithm, enhancing the precision of the positioning system.
• Developed a user-friendly GUI to control and monitor the system, improving user engagement and system manageability.

Responsibilities include: Using MAML, DRL, and other AI techniques in Human-Robot Collaboration

• Project 1: Meta-reinforcement Learning Based Control Approach for Assisting HRC in Personalized Production (2021/12-2022/08)
• Keywords: Model diagnostic meta-learning, Proximal policy optimization DRL, Robotics Perception
• • Awarded paper by HKIE CA Paper Competition 2021/22
• • Designed self-learning robotic assisted systems (SLRAS) with self-adapt DRL algorithms to assist human operators
• • Enhanced the performance by about 75% on average
• Project 2: Mutual Cognitive Human-Robot Collaborative Manufacturing System (2022/01-2023/02)
• Keywords: Deep Reinforcement Learning, Augmented Reality, HRC
• • Awarded HKICT Student Innovation Grand Award and Gold Award and APICTA Award
• • Integrated virtual and realistic task planning and dynamic guidance of visual data to enhance the intelligence of robots and to facilitate information sharing with Augmented Reality
• • Conducted human-in-the-loop control approaches