Tianhao Li AI for Science Scholar

AI for Science · Scientific Machine Learning

Tianhao Li

Machine learning for scientific discovery.

I build data-efficient ML systems for expensive scientific search—combining active learning, surrogate models, model evaluation, and LLM-assisted workflows toward closed-loop discovery.

Email Google Scholar GitHub LinkedIn
Job
Applied Scientist II Intern Amazon Prime Air
Education
Ph.D. Candidate Johns Hopkins University
Award
Amazon AI PhD Fellow 2025–2027
Tianhao Li outdoors against a mountain landscape

Selected work

ML systems

Projects across data-efficient adaptation, candidate ranking, and LLM-assisted scientific data systems.

Data-efficient adaptation · MLIPs · 2024–present

Adapting scientific models with targeted data

Similarity-guided data selection fine-tunes ML interatomic potentials on compact, relevant training sets, reducing reliance on new high-fidelity calculations.

Workflow connecting disordered structures, interatomic potentials, and energy-based formability descriptors

Surrogate modeling · Candidate ranking · 2025

Ranking candidates in large design spaces

Random-forest surrogates built on physically motivated disorder descriptors rank more than 20,000 candidates under activity, stability, and supply constraints.

High-entropy alloy screening workflow and candidate design space

LLM-assisted retrieval · Scientific data · 2023–present

Making scientific data usable by models and people

A tool-enabled interface translates natural-language questions into structured retrievals, supporting analysis across large scientific simulation collections.

Element-frequency maps across the CHAOS high-entropy oxide database

Research record

Selected papers

Current work in scientific model evaluation, data-driven discovery, and simulation.

  1. Revised manuscript

    Structural Benchmarks Overlook Density-of-States Errors in Machine-Learned Interatomic Potentials

  2. Precision Chemistry

    Stabilizing Iodine in Pyrochlore: Toward New Nuclear Waste Forms

  3. Nano Futures

    The Search for High-Entropy Fuel-Cell Catalysts Using Disorder Descriptors

  4. npj Computational Materials

    High Entropy Powering Green Energy: Hydrogen, Batteries, Electronics, and Catalysis

  5. Materials Horizons

    Beyond the Four Core Effects: Revisiting Thermoelectrics with a High-Entropy Design

  6. Frontiers in Chemistry

    Effects of H2–H3 Phase Transition Reversibility on Ni-Rich Layered Oxide Cathodes

Full list on Google Scholar · ORCID

Background

Education

Johns Hopkins University

Ph.D. · Materials Science & Engineering

Advisor: Corey Oses

Duke University

M.S. · Materials Science & Engineering

Changsha University of Science and Technology

B.S. · Materials Science & Engineering

Academic work

Teaching Python & computational modeling Service Peer review, guest editing & research mentorship

Contact

Research & collaboration.

I’m happy to discuss AI for Science, data-efficient learning, model evaluation, scientific data systems, and applied research.

tli114@jh.edu Scholar GitHub LinkedIn