I'm a PhD researcher who trains prediction models, adapts existing models to specific tasks, and builds tools that answer questions from documents. My experience includes demand forecasting for retail and contract analysis covering more than 15 types of clauses. I work across data preparation, model evaluation, and the software that brings these pieces together. For contract review, that also meant linking answers to the original text so people could check the evidence themselves.
A selection of work spanning ML, NLP, and full-stack data applications.
Published work in AI and biomedical NLP.
Introduces MetaphorPrompt2, which improves molecular regulatory pathway extraction from biomedical literature by emphasizing structural relations and functional roles of biological entities rather than surface-level grammar. Integrates five components that reduce parsing complexity and mitigate error propagation, achieving a 31% improvement in edge-prediction F1 over the no-metaphor baseline.
Parth Patel, Yu-Chiao Chiu, Yufei Huang, Jianqiu Zhang
End-to-end framework using GPT-4 and In-Context Learning to extract hierarchical Molecular Regulatory Pathways from biomedical literature. Constructed m6A-KG — a knowledge graph with 2.4K nodes and 4.7K edges. Designed G-Eval, an LLM-based annotation-free evaluation framework.
Introduces MetaphorPrompt, a novel approach using GPT-4 to generate metaphors that map biological processes onto real-world scenarios, improving extraction of molecular regulatory pathways from biomedical literature. Tested on the reguloGPT dataset; achieves improved precision, recall, and F1 scores in causal event link prediction via analogical reasoning and in-context learning.
Parth Patel, Yu-Chiao Chiu, Yufei Huang, Jianqiu Zhang
University of Texas at San Antonio
University of Texas at San Antonio
I'm open to research collaborations, full-time data science and ML engineering roles, and PhD internship opportunities. Feel free to reach out.