Gaoyuan / KRATSZ

I build AI systems for biology and laboratory automation.

Graduate student at the University of Chinese Academy of Sciences / SIAT, working on AI-assisted protocol generation, protein engineering assessment, and agentic tools for scientific workflows.

Base
Shenzhen, China
Bridge
Biology x AI x automation
Mode
Research systems, benchmarks, tools

Current direction

Making scientific workflows easier to design, run, and evaluate.

My work sits between wet-lab automation and AI agents: turning lab intent into executable protocols, checking whether generated work is actually valid, and building evaluation loops that make scientific automation more reliable.

Featured projects

Selected work with a clear research thread.

Protein engineering 02

CAPE

Critical assessment work for protein engineering, connecting biological design problems with evaluation methods that can support stronger model claims.

Scientific tooling 03

labscriptai-ot

Practical tooling around Opentrons and LabScript-AI workflows, including integration surfaces that help agentic systems meet real lab constraints.

Research focus

Three areas that reinforce each other.

A

AI for lab automation

Converting experimental goals into executable protocols while keeping equipment, safety, and validation in the loop.

B

Scientific agent workflows

Building agent systems that can plan, call tools, check outputs, and leave useful evidence instead of only producing text.

C

Protein engineering assessment

Studying evaluation problems in biological design where reliable comparisons matter as much as new model ideas.

Selected updates

A lightweight place for public progress.

LabScript-AI authoring and evaluation

Developing benchmark-backed workflows for AI-generated lab protocols and practical Opentrons automation.

Scientific agents and local tools

Exploring tool-using agents for research tasks, local automation, and reproducible scientific work.

Biology and model evaluation

Connecting biological questions with careful evaluation design across protein engineering and related systems.

Contact

Open to research conversations and useful collaborations.