cfDNA-Scientist: Multi-Agent Biomedical Feature Discovery
A self-evolving multi-agent system that proposes, implements, and evaluates biomedical feature-extraction code. Reaches a cross-validated AUC of 0.916 on breast cancer detection.
•
1 min read
Po-Chung (Bryan) Hsieh
M.S. Student in Electrical and Computer Engineering
Professional Summary
I am an M.S. student in Electrical and Computer Engineering at UC San Diego, with a B.S. in Electrical Engineering from National Taiwan University. I build AI agents and reinforcement learning methods for scientific discovery, including multi-agent systems for biomedical feature discovery and policy optimization for discrete generative models, and I previously worked as an LLM engineer intern at the Inventec Artificial Intelligence Center.
Education
M.S. in Electrical and Computer Engineering
2026-09-01
2028-06-01
University of California San Diego, USA
B.S. in Electrical Engineering
2021-09-01
2026-06-01
National Taiwan University, Taiwan
Interests
AI Agents
Reinforcement Learning
Machine Learning for Biomedical Discovery
Research Interests
I am interested in building AI agents that learn from feedback and support scientific discovery. My recent work spans policy optimization for discrete flow matching and agent workflows for biomedical feature discovery.
Selected Projects
DoMinO: Discrete Flow-Matching Policy Optimization
A reinforcement learning framework for fine-tuning discrete flow-matching models. Outperforms prior state-of-the-art by 10.6% on DNA design and 15.4% on protein design.
•
1 min read
Publications
(2026).
Discrete Flow Matching Policy Optimization.
arXiv preprint arXiv:2604.06491.
(2026).
The Binding Effect: Analyzing How Multi-Dimensional Cues Form Gender Bias in Instruction TTS.
arXiv preprint arXiv:2603.20743; accepted to INTERSPEECH 2026.
(2025).
Dual-Process Scaffold Reasoning for Enhancing LLM Code Debugging.
arXiv preprint arXiv:2511.08052.