• Built cfDNA-Scientist, a self-evolving multi-agent system that autonomously proposes, implements, and evaluates biomedical feature-extraction code in executable research environments.
• Designed a preliminary literature-search workflow to identify established cfDNA fragmentomic features, build a feature catalog, and validate baseline performance.
• Evaluated cfDNA-Scientist on breast cancer detection, achieving a cross-validated AUC of 0.916 compared with the baseline’s 0.831 under the same evaluation protocol.

• Co-developed DoMinO, a reinforcement learning framework that formulates discrete flow-matching inference as an inner Markov decision process, enabling exact trajectory log-likelihood optimization.
• Implemented REINFORCE- and PPO-based fine-tuning methods with cross-entropy and generalized KL regularization to mitigate reward hacking and preserve sequence naturalness.
• Evaluated DoMinO on regulatory DNA and protein design, outperforming prior state-of-the-art baselines by 10.6% on DNA predicted activity and 15.4% on protein pred-ddG.

Taipei, Taiwan.

• Proposed a dual-process scaffold reasoning framework for LLM code debugging, achieving an 88.91% pass rate on DebugBench and outperforming prior prompting methods.
• Built a NeMoClaw-based agent backend that automated thermal-design exploration and component-placement optimization for 30 notebook system-design teams.
• Developed an automated circuit-validation system supporting 20 antenna engineers, enabling design verification without specialized circuit-design software.