Skip to content
Jingyi Li

Machine learning for scientific discovery

Jingyi Li

Machine Learning Scientist

Ph.D. in Computational Biophysics
Penn State, 2026

I develop machine learning methods for scientific problems where geometry, structure, and physical constraints matter.

My work combines geometric deep learning, generative modeling, molecular simulation, and graph algorithms. I approach these problems as complete scientific systems: represent the object carefully, build the model and data pipeline around that representation, then validate the result against physical and domain-specific evidence.

My recent research focuses on RNA structure. I am the first author of ChironRNA, an equivariant diffusion system for resolving steric clashes, and ATLAS, a graph-based library for searching structural RNA motifs.

jingyili.mechanics@gmail.comResumeLinkedIn

01

Research

My foundation is in mechanics and simulation. The scientific domain may change, but the method remains consistent.

Geometric learning

Representing coordinates, interactions, symmetries, and physical constraints before choosing the learning architecture.

Scientific systems

Building the data, training, generation, diagnostics, and benchmarking pipeline around a model, primarily with Python and PyTorch.

Mechanistic validation

Testing learned outputs against geometry, simulation, and domain baselines so that scientific claims survive inspection.

02

Selected research

2026First author

ChironRNA

Steric clash resolution in RNA structures via E(3)-equivariant diffusion

A hierarchical diffusion system that identifies damaged regions with MolProbity diagnostics and regenerates them while preserving valid molecular geometry. The system reduced steric clashes by 80% on more than 80% of the test set.

Interactive research storyPreprint

2026First author

ATLAS

A graph-based 3D RNA motif library

An automated pipeline that processed 8,791 experimental RNA structures into atomic coordinates and graph representations, making 433,996 motifs searchable through graph isomorphism.

Source codePreprint

03

Selected publications

  1. ChironRNA: Steric Clashes Resolution in RNA Structures via E(3)-Equivariant Diffusion

    Jingyi Li, Jian Wang, and Nikolay V. Dokholyan

    bioRxiv, 2026 · First author

  2. ATLAS: Graph-based 3D RNA Motif Library Incorporating non-Watson-Crick Interactions

    Jingyi Li, Jian Wang, Srinivasan Ekambaram, and Nikolay V. Dokholyan

    bioRxiv, 2026 · First author

  3. Diffusion of water nanodroplets on graphene with double-vacancy

    Lei Deng, Jingyi Li, Shan Tang, and Zhen Guo

    Applied Surface Science 573, 151235, 2021

04

Elsewhere on this site

The research profile is the front door. These sections retain the more exploratory work that has always been part of this site.

ReadingInteractive explanations of papers, books, and technical topics.PhotosEdited photographic issues from trips, places, and events.ToolsSmall browser tools, including direct peer-to-peer transfer.EventsA monthly guide to events near Hershey, Pennsylvania.

05

Scientific AI practice

Jingyi Li Scientific AI LLC is my independent practice for research teams. It focuses on scientific modeling, research infrastructure, and automation with explicit validation.

Visit the practice page →