# Cady Tianyu Xu - Personal Website > This is the personal academic portfolio of Cady Tianyu Xu, a researcher at Google DeepMind working on LLM agents and multimodal AI. ## About Cady Xu Cady Tianyu Xu is a researcher at Google DeepMind. Her work focuses on multimodal LLM agents that move beyond open-ended generation toward closed-loop execution, bridging the gap between high-level reasoning and autonomous task completion in real-world environments. Previously, she was a machine learning engineer on the Google XR team, developing perception models and multimodal LLMs for immersive, context-aware interaction. Her recent work spans multimodal human-centered systems and interactive AI, with publications at UIST and CHI. She holds a Bachelor's degree in Computer Science and Political Science from UC Berkeley. ## Research Interests - Large Language Model (LLM) Agents - Multimodal AI and Machine Learning - Extended Reality (XR) and Spatial Computing - Audio-Visual Perception and Interaction - Human-Computer Interaction (HCI) ## Selected Publications ### MoXaRt (CHI 2026) Audio-Visual Object-Guided Sound Interaction for XR. A real-time XR system that uses audio-visual cues to separate sound sources and enable fine-grained sound interaction. Authors: Tianyu Xu et al. ### SAMOSA (UIST 2025) Enhancing XR Auditory Realism via Multimodal Scene-Aware Acoustic Rendering. An on-device system that renders spatially accurate sound by dynamically adapting to physical environments. Authors: Tianyu Xu et al. ### EI-Lite (UIST 2025) Electrical Impedance Sensing for Micro-gesture Recognition and Pinch Force Estimation. A lightweight wrist-worn device for micro-gesture recognition. Authors: Junyi Zhu, Tianyu Xu et al. ### Steerable Chatbots (arXiv 2025) Personalizing LLMs with Preference-Based Activation Steering. Lightweight LLM personalization via activation steering. Authors: Jessica Y. Bo, Tianyu Xu et al. ### CaliPSo (ICML 2025 Workshop) Calibrated Predictive Models with Sharpness as Loss Function. Neural network model for uncertainty quantification with marginal calibration. Authors: Alexandre Capone, Kamron Zaidi, Tianyu Xu et al. ### Liquid EIT (ICRA 2024) Liquids Identification and Manipulation via Digitally Fabricated Impedance Sensors. Robot sensing for liquid manipulation. Authors: Junyi Zhu, Young Joong Lee, Yiyue Luo, Tianyu Xu et al. ## Contact Information - Full Name: Cady Tianyu Xu - Current Role: researcher at Google DeepMind - Email: tyx@google.com - LinkedIn: https://www.linkedin.com/in/cadyxu/ - Google Scholar: https://scholar.google.com/citations?user=nx1V8M4AAAAJ - ORCID: https://orcid.org/0009-0009-9135-6080 - ACM Profile: https://dl.acm.org/profile/99661715423 - Work Page: https://research.google/people/cadyxu/ - Website: https://cadyxu.com ## Site Links - Home: https://cadyxu.com/ - Publications: https://cadyxu.com/publications/ - CV: https://cadyxu.com/cv/ - Projects: https://cadyxu.com/projects/