Yichen Zhang · Researcher

Building generative systems for scientific discovery.

Research Student at Osaka University, working toward generative and foundation-model approaches for molecules and biological sequences.

Osaka, Japan AI × Life Science
Yichen Zhang at the Golden Gate Bridge
San Francisco, 2026
Scroll
A

Generative AI for Drug Discovery

How can generative models navigate molecular and biological design spaces while respecting scientific constraints?

B

Multimodal Intelligence

How can models connect visual evidence, language, and structured outputs with dependable reasoning?

C

Embodied AI & Robotics

How can agents ground perception and language in actions that remain robust outside curated environments?

01

VLM Fine-tuning for Schema-Constrained Food Extraction

2025

Fine-tuned SmolVLM2-500M for reliable JSON extraction from food images and released reproducible weights for downstream integration.

02

GeneLM-Evo2: Zero-shot Genomic Variant Pathogenicity Analysis

2025

Built a zero-shot pathogenicity scoring pipeline around Evo2 7B with long-context genomic inference and a scalable GPU serving stack.

03

Efficiency Analysis of Tiny Recursive Models for Reasoning

2025

Reproduced Tiny Recursive Models and measured how accuracy, throughput, and compute budgets interact on Sudoku-Extreme reasoning tasks.

  • Joined Osaka University as a Research Student, preparing for Ph.D. study in generative AI for drug discovery.
  • Completed an M.S. in Computer Engineering at New York University.
  • Open-sourced GeneLM-Evo2 and a schema-constrained VLM fine-tuning pipeline.

Collaboration

Interested in generative models for science?

I am open to research conversations, collaborations, and thoughtful exchanges across AI and the life sciences.

Start a conversation