MAO CHI HE

MAO CHI HE (Matthew)

Hi, I’m Mao-Chi He, a graduate of the Interdisciplinary Program of Engineering at National Tsing Hua University, where I studied both Mechanical Engineering and Computer Science. During my senior year, I joined the University of California, Berkeley as an exchange student in the Department of Computer Science, further strengthening my background in AI, software development, and computational thinking.

I have always been fascinated by robotics and artificial intelligence, especially their potential to solve real-world problems and improve people’s daily lives. Through my internship and research experiences, I have gained hands-on experience in implementing the latest AI technologies and applying them to practical robotic systems. These experiences have also helped me identify meaningful challenges in the robotics field and motivated me to develop solutions that bridge AI, software, and physical systems.

Email / CV / Github / LinkedIn

Education

	University of California, Berkeley Aug 2024 - May 2025 Exchange Student, Computer Science Department Location: Berkeley, CA
	National Tsing Hua University (Taiwan) Aug 2021 - May 2024 B.S., Interdisciplinary Program of Engineering (CS + MechE) Location: Hsinchu, Taiwan

Experience

	Microsoft Sep 2025 - Mar 2026 Research & Development Intern @ MAI-A Ads Taipei, Taiwan Architected and developed RecapRon, an AI-powered meeting recap platform that converts long recordings into concise highlight videos (~10% of original length). Engineered LLM pipelines for transcript segmentation, summarization, and script generation, improving output coherence and downstream reliability. Designed reusable agent-native workflow modules with Claude Code Skills to decompose complex AI tasks into scalable components.
	Institute of Information Science, Academia Sinica Jul 2025 - Mar 2026 Research Assistant Taipei, Taiwan Developed an event-based optical tactile sensing framework for robotic manipulation with real-time contact, compliance, and slip perception at 1500 FPS. Proposed a self-supervised spatio-temporal pretraining method for tactile data to improve representation learning and generalization across downstream tasks. Built a real-time multimodal robotics pipeline by integrating the tactile sensor, gripper, robot arm, and VLA model through ROS2.
	CVLab, National Tsing Hua University Jan 2025 - Dec 2025 Undergrad Research Assistant Hsinchu, Taiwan Applied 3D Gaussian Splatting to real-time rendering and 3D scene reconstruction, improving simulation efficiency and geometric fidelity. Constructed digital twin models from sparse-view inputs (e.g., 4 images) for real-world scene simulation.
	CNELAB, National Tsing Hua University Aug 2023 - Dec 2024 Undergrad Research Assistant Hsinchu, Taiwan Designed and programmed a frequency flashing device using Arduino UNO R3 with a 9-LED array for SSVEP experiments. Collected 32-channel EEG signals and applied signal processing and machine learning methods, including CNN, CCA, SVM, and Random Forest.

Publications

Patterson Hsieh, Chia-Jui Yeh*, Mao-Chi He, Wen-Han Hsieh, Elvis Hsieh†. "Seg the HAB: Language-Guided Geospatial Algae Bloom Reasoning and Segmentation." NeurIPS 2025 Workshop on Tackling Climate Change with Machine Learning, 2025. [Paper]
*Equal contribution †Co-supervising authors

Projects

AI Agent Project Jan 2025 - Present

Developer (Berkeley, CA)

Designed a LangGraph + LangChain + Qdrant workflow to autonomously draft replies for medical insurance e-mails.
Targeted a 50% reduction in insurance paperwork time, potentially saving physicians 7-10 hours per week for patient care.

AI Cup Hackathon Oct 2024 - Nov 2024

Participant (7th/487 teams) - Taiwan

Applied OCR to extract text from images and generate embeddings for retrieval and analysis.
Integrated ChatGPT, Gemini, Claude, and Llama APIs to improve answer quality for image-based question answering.

GenAI Stars Hackathon May 2024 - Jul 2024

Participant (Semifinalist & Merit Award) - Taiwan

Built a sports-analysis pipeline using YOLOv8 for player detection, KMeans for team classification, and Whisper for speech-to-text.
Integrated video processing, speech understanding, and GPT-4-based summarization into a Flutter application.