cfDNA-Scientist: Multi-Agent Biomedical Feature Discovery

Mon, 01 Jun 2026 00:00:00 +0000

Role: Project Lead · Northwestern University (Remote) · Jun 2026 – Oct 2026

cfDNA-Scientist is a self-evolving multi-agent system that autonomously proposes, implements, and evaluates biomedical feature-extraction code in executable research environments, using cell-free DNA data.

What I did

Built the multi-agent system. Agents coordinate hypothesis generation, feature implementation, verification, critique, and curation, running the generated code in executable research environments.
Designed a literature-search workflow. A preliminary workflow identifies established cfDNA fragmentomic features, builds a feature catalog, and validates baseline performance.
Evaluated on breast cancer detection. cfDNA-Scientist reached a cross-validated AUC of 0.916, compared with 0.831 for the baseline under the same evaluation protocol.

See the for how this fits into my broader work.

DoMinO: Discrete Flow-Matching Policy Optimization

Wed, 01 Apr 2026 00:00:00 +0000

Role: Project Co-Lead · Northwestern University (Remote) · Feb 2026 – Apr 2026

DoMinO is a reinforcement learning framework for fine-tuning discrete flow-matching models toward a reward.

What I did

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

Paper: . See also the .

Projects | Po-Chung (Bryan) Hsieh

cfDNA-Scientist: Multi-Agent Biomedical Feature Discovery

What I did

DoMinO: Discrete Flow-Matching Policy Optimization

What I did