Xubing Hao

About Xubing Hao

I am an Instructor in the Department of Neurology at McGovern Medical School (UTHealth Houston), working at the intersection of clinical AI, knowledge representation, and large language models (LLMs). My work focuses on turning complex, messy biomedical data into representations that support reasoning, interpretability, and real-world clinical use.I actively build and evaluate applied clinical AI systems. This includes fine-tuning large language models for EEG impression summarization, designing retrieval-augmented generation (RAG) for Alzheimer\'s Disease (AD) data element mapping, and developing ontology-powered tools that help domain experts interact with structured knowledge more effectively. I enjoy working across the full stack: from theoretical definitions and proofs, to Python/Java implementations, to user-facing evaluation with clinicians.

Experience

1. Instructor

   Mcgovern Medical School

   Jun 2025 — Present · Houston, TX, US

   AI-Driven Clinical Impression Summarization (EpiToMe EMR)1. Design and deploy a clinician-customizable clinical impression summarization system within the EpiToMe epilepsy Electronic Medical Record (EMR)2. Adapt large language models (LLMs) via parameter-efficient fine-tuning (LoRA/PEFT) and reinforcement learning (RL)–based post-training (vLLM) to align model behavior with clinician preferences and standardsUsing Real-World Data to Derive Common Data Elements for AD and ADRD Research (ReCARDO)1. Built hybrid retrieval system to align heterogeneous Alzheimer’s Disease (AD) data elements across large-scale real-world datasets, including EHRs, claims data, and multi-institutional research cohorts2. Implement multi-stage workflows using Elasticsearch (BM25), retrieval augmented generation (RAG), and LLM-based semantic reranking, improving recall while maintaining precision for clinical evaluation

Education

• UTHealth Houston

  Doctor of Philosophy - PhD, Biomedical Informatics

• Jilin University

  Bachelor's degree, Software Engineering