Novalith — Follow the Evidence. Advance Science.
We bring feedback from computation and experiments into AI training, and use the data we gather to develop domain models and training environments for biomedical R&D.
Services & Collaboration
Research Practice
Capture how researchers decide
We document how researchers interpret results, weigh evidence, and change course so AI can learn from their decisions.
Charles Darwin, Notebook B, 1837
Training Environments
Build environments for testing
We build reinforcement learning environments that turn computational and experimental results into faster feedback for training.
Ludwig’s kymograph, 19th-century print
Domain Models
Support domain model development
We support the development of models that help biomedical researchers interpret results and decide what to investigate next.
Santiago Ramón y Cajal, 1899
Research Partnerships
Test research ideas together
We work with researchers and R&D teams to turn a question worth pursuing into an idea we can test together.
Lavoisier’s lab, after Mme Lavoisier
The team behind Novalith
We’re scientists, AI researchers, and engineers. We work with research partners in the life sciences to shorten the path from a research question to a tested result.
Xi Liu
Founder · Scientist
PhD, Peking University. Postdoc, University of Pennsylvania. Led AI projects at a publicly traded medical device group.
Junhui Ma
Co-founder · Builder
Previously led strategy products at ByteDance and Guangnian Zhiwai.
Shiqi Yang
Co-founder · Curator
Bachelor’s, Peking University. Master’s, University of Pennsylvania. Led LLM data annotation at ByteDance.
Anna Zhang
Scientist
Darren Sun
Builder
Felix Wang
Builder
Cris Xu
Connector
Leslie
Shaper
Koutian
Curator
Work with Novalith
From a question worth testing to a system that works, we’re looking to build alongside researchers, R&D teams, and technology partners.
Research partnerships
Tell us about a research question, an experimental workflow, or a lab instrument you’d like to connect to an AI system.
Discuss a collaborationCareers
If you work on AI agents, model training, robotics, or experimental science, we’d like to hear from you.
Join Novalith
Follow the Evidence. Advance Science.
元石银河 — 为科学探索,循迹向前。
我们让计算与实验的反馈进入模型训练,并基于积累的专业数据,发展面向生物医学研发的领域模型与训练环境。
服务与合作
科研经验
沉淀科研判断经验
记录研究者如何解读结果、权衡证据和调整计划,让 AI 从这些判断中学习。
查尔斯·达尔文,笔记 B,1837
训练环境
构建科研训练环境
我们构建面向科研的强化学习环境,让 AI 更快获得计算与实验反馈,加快训练迭代。
卡尔·路德维希的记纹鼓,19 世纪版画
领域模型
支持领域模型开发
支持面向生物医学研究的模型开发,帮助研究者解读结果、判断下一步该研究什么。
圣地亚哥·拉蒙-卡哈尔,1899
科研合作
共同检验科研想法
从一个值得追问的问题开始,与研究者和研发团队一起选择方法、检验想法。
拉瓦锡的实验室,据拉瓦锡夫人素描
认识元石银河团队
我们是科学家、AI 研究者和工程师。我们与生命科学领域的科研伙伴一起,缩短从提出问题到实验验证的路径。
刘曦
Founder · Scientist
北京大学理学博士、宾夕法尼亚大学博士后,曾负责上市医疗器械集团 AI 项目。
马俊辉
Co-founder · Builder
前字节跳动、光年之外策略产品负责人。
杨诗琦
Co-founder · Curator
北京大学本科、宾夕法尼亚大学硕士,前字节跳动大模型数据标注负责人。
Anna Zhang
Scientist
Darren Sun
Builder
Felix Wang
Builder
Cris Xu
Connector
Leslie
Shaper
Koutian
Curator
与元石银河一起工作
从一个值得验证的问题,到一套能够运行的系统,我们期待与研究者、研发团队和技术伙伴一起探索。
为科学探索,循迹向前。