The Hong Kong University of Science and Technology
PhD in Computer Science and Engineering
Advisor: Prof. Qian Zhang
Multimodal learning, intelligent agents, and AI for health.
Curriculum vitae
Xuange Liang
PhD student in Computer Science and Engineering · HKUST
Research interests
Education
Columbia University
MPH in Biostatistics
Outstanding Student · CEOR
Peking University
BS in Intelligence Science and Technology
Publication
AAAI-26 Workshop on Health Intelligence (W3PHIAI-26)
2026 · Short paper
Research & projects
HAEnv
An open-source pipeline for generating synthetic patient histories and evaluating health agents against answers derived from the simulation. An interactive demo makes the patient timeline and evaluation process visible.
Developed patient-generation and evaluation tooling, verification checks, and the interactive demonstration as part of my work at Theta Health.
Learning treatment effects from EHRs
Adapting pretrained electronic health record representations to estimate heterogeneous treatment effects in small clinical cohorts.
Combined pretrained CLMBR embeddings, dimensionality reduction, and an R-learner to study treatment-effect estimation across clinical outcomes.
Health Agent Sandbox
Designing a simulation environment for evaluating health agents’ long-term decisions, with reproducible patient trajectories and counterfactual comparisons.
Research design and evaluation specification.
Multimodal 3D Object Detection
Research with Prof. Yongtao Wang. Developed continuous-depth components using Lift-Splat-Shoot and ZoeDepth, with a detection head and data pipeline in MMDetection 3D.
Experience
AI Intern · Shanda Group
- Worked on LLM agents using wearable and health-examination data.
- Built data-processing and synthetic-data workflows for longitudinal health records.
Back-End Engineer Intern · JD Tech
- Developed Go APIs and scripts for database management and migration.
- Worked with Docker, Kubernetes, and MySQL query optimization.
Data Scientist Intern · CloudWalk Technology
- Integrated and standardized health-related records from multiple tables.
- Built and evaluated models for hotel health-risk prediction.
Tools & methods
Python, PyTorch, R, SQL, Go, Git, Docker, Machine learning, Causal inference, Agent evaluation, Interactive visualization.