D-IFSC is an interactive sport climbing data platform designed and developed using HTML, CSS, and JavaScript to enhance usability and engagement for both novice and expert users. It integrates large-scale competition and athlete data extracted and processed from the IFSC website via Python and Selenium, and features domain-specific visualizations—including radar, chord, and bubble charts—optimized through iterative user research to support intuitive analysis, improve comprehension, and accommodate users with diverse backgrounds.

UnIEM (Unified Interactive Environment Model) is a novel interaction model designed to enable real-time reconfiguration of Distributed User Interfaces (DUIs) across multiple IoT devices in dynamic environments. The system is built on a modular framework featuring custom UI component mapping and a central agent system (DuiAgent) that manages, distributes, and adapts interface elements based on user preferences, accessibility needs, and contextual factors. The model was implemented through an application on HarmonyOS and validated via a multi-device video migration app in an educational setting, demonstrating inclusive, user-driven interface customization and showcasing expertise in interaction design, smart IoT ecosystems, and user-centered development.

Digital Corpse Donation is a Caring Digital Burial Process. The data trails we leave online form a digital twin, reflecting our personality, memories, and works. While our physical lives end, our data can endure, sparking discussions on digital death, privacy, and time. Humans have rituals for physical death—how can we create digital ones? In collaboration with Axa Health, experts in after-death data consulting, we propose the Digital Corpse Donator, an interactive experience promoting data donation. This encourages preserving meaningful data for institutions while discarding what's unnecessary, fostering awareness of data management, better personal practices, and the role of companies in shaping our digital legacies.

NewsieAI is a dual-agent platform allows news retrieval through intelligent profile management and context-aware information gathering. Built on LangChain and Model Context Protocol (MCP), it enables natural language profile building and introduces a new distributed agent payment ecology through HTTP 402 standardization, allowing LLMs to own digital assets and operate autonomously on blockchain networks.

SCROLL TO UNLOCK 0%

Developed multiple LLM-powered Discord bots tailored for distinct use cases. Key projects include: a role-playing bot with real-time persona adaptation and emotional support features, built using dual-model architecture and context engineering for long-term memory; and a game strategy assistant leveraging RAG with Sentence-Transformer and ChromaDB for accurate retrieval of game mechanics. Optimizations such as LoRA fine-tuning, async prompt distillation, and re-ranking techniques reduced API costs by 20–30% and improved response relevance.

Digital Corpse Donation is a Caring Digital Burial Process. The data trails we leave online form a digital twin, reflecting our personality, memories, and works. While our physical lives end, our data can endure, sparking discussions on digital death, privacy, and time. Humans have rituals for physical death—how can we create digital ones? In collaboration with Axa Health, experts in after-death data consulting, we propose the Digital Corpse Donator, an interactive experience promoting data donation. This encourages preserving meaningful data for institutions while discarding what's unnecessary, fostering awareness of data management, better personal practices, and the role of companies in shaping our digital legacies.

SCROLL TO UNLOCK 0%

Lumen is an intelligent health platform integrating light exposure tracking, adaptive light intervention, and biometric monitoring, specifically designed for Seasonal Affective Disorder (SAD) patients and indoor professionals. Integrating and analysis data from native health application and Lumen enables holistic health management from environmental sensing to mind-body regulation through visualized light analytics and user-centric photoperiod rhythm optimization.

SCROLL TO UNLOCK 0%

Maple Valley is a immersive Virtual Reality (VR) game, developed using Unity 3D and C++ scripting, emphasizes exploration, gathering, crafting, and building in an interactive environment. It features a VR-adaptive interaction system designed for seamless engagement with the inventory, utilizing location-based detection aligned with intuitive human behavior to ensure natural and effortless gameplay.

OneClick is a lightweight Android application designed to provide efficient, customizable access to apps and actions through dynamic shortcuts. Built with an object-oriented architecture, the system integrates gesture recognition, floating windows, and a real-time shortcut management system powered by a purpose-built data structure for efficient storage and retrieval. The app’s UI/UX was refined through comprehensive quantitative and qualitative usability research, ensuring simplicity and minimal cognitive load for users.

This project involved designing and implementing a custom TCP-based file synchronization protocol in Python to enable Large Efficient Fast and Trustworthy file sharing between two PCs. Built using Object-Oriented Programming (OOP), the system is structured around three key components—File Scanner, Listener, and Downloader—each with specialized functions. A custom header data structure supports file splitting and error detection, allowing large files to be broken into manageable chunks for optimized memory usage, transmission efficiency, and prevention of issues like infinite loops or premature overwriting during transfer.

Torch is a vibration-feedback guiding glove designed to translate distance information into haptic signals, enabling blind individuals to navigate indoor environments more independently. The design was grounded in accessibility research, incorporating user interviews and behavioral analysis to uncover key pain points. As we identified the gap between the design with prominent visibility in existing assistive devices and users with a high need for self-independence through interviews. Implemented using Arduino, the glove was tested and quantitatively evaluated to enhance user experience and effectiveness in real-world indoor scenarios. Enhanced user autonomy and dignity while ensuring safety in emergency situations.

I use a camera to record what I see, capture beautiful moments of my life.