Haotong Liang

Ph.D., Materials Science & Engineering hliang16@umd.edu  ·  auroralht.github.io  ·  College Park, MD

Autonomous experimentation for materials discovery — I build the deep-learning and Bayesian-inference workflows that let instruments run themselves. I applied Bayesian optimization to rapidly identify robust growth conditions for metastable materials, improving process reliability and yield in semiconductor-style manufacturing environments. My goal is to raise scientific throughput by one to two orders of magnitude through self-driving labs. I work across the full stack, from low-level hardware control to machine-learning pipelines, and pick up new tools, techniques, and domains quickly.

Machine Learning
Deep learning / CNN / GAN / GNN / LLM / Bayesian optimization / Gaussian processes / PyTorch
Programming
Python / TypeScript / JavaScript / Java / C / SQL / MATLAB / LabVIEW
Lab & Instruments
PLD / RHEED / XRD / SEM-EDS / TEM-EELS / AFM / WDS
Systems & Tooling
React / message queues / computer networking / Linux / shell / web scraping / SolidWorks

Education

Ph.D., Materials Science & Engineering
University of Maryland, College Park
Jan 2021 – 2026
B.S., Materials Science & Engineering
University of Maryland, College Park · Minor in Computer Science
Oct 2016 – Jan 2021

Research Experience

Takeuchi Research Group · University of Maryland
Research Assistant
Oct 2018 – 2026
  • Years of hands-on thin-film fabrication using pulsed-laser deposition (PLD); proficient in XRD characterization across sputtered combinatorial libraries, epitaxial single-crystal films, and polished metal surfaces.
  • Employs AFM, SEM-EDS, and WDS for multi-scale surface morphology and composition analysis; skilled in TEM-based nanocrystalline characterization including structural topology, phase identification, elemental mapping (EDS), and valence-state analysis (EELS).
  • Proven cross-disciplinary collaborator with a strong track record of joint projects across multiple research groups and institutions.
  • Conducted machine-learning–driven research for materials discovery across phase-change materials, antiferromagnets, and high-Tc superconductors.
  • Developed CRYSPNET, a neural network that predicts crystal structures from chemical compositions; the learned latent space serves as a robust materials descriptor.
  • Collaborated with NIST researchers to build a generative adversarial network (GAN) that significantly improved structure-classification accuracy for neutron diffraction data.
  • Built a Cascade Mask R-CNN instance-segmentation model for RHEED images, enabling automated quantitative analysis and phase mapping — the foundation of ongoing autonomous experimentation workflows.
Haotong Liang · Résumé 1 / 2
Haotong Liang hliang16@umd.edu · auroralht.github.io

Research Experience continued

  • Developed a Bayesian optimization framework for direct energy deposition and demonstrated the workflow on MAR-M247.
  • Built an end-to-end image-analysis pipeline for autonomous pulsed-laser deposition experiments, including a user-friendly interface designed for seamless handoff to collaborators.
  • Applied Bayesian optimization to discover optimal growth conditions for metastable hTbFeO3 and ferroelectric HZO, demonstrating autonomous synthesis of technologically relevant oxide thin films.
  • Developed an agentic RAG system to assist with scientific research questions — the system autonomously navigates peer-reviewed literature using tools from multiple providers, synthesizes findings across papers, and draws evidence-based conclusions.

Publications nine works, peer-reviewed & preprint

Self-driving thin film laboratory: autonomous epitaxial atomic-layer synthesis via real-time computer vision analysis of electron diffraction
H. Liang, Y. Sun, R. Paxson, C.-Y. Lee, A. T. Hall, Z. Warecki, J. Cumings, H. Koinuma, A. G. Kusne, M. Lippmaa, I. Takeuchi. arXiv:2602.20432 (2026).
Mechanism-resolved PFM of ferroionic and ferroelectric responses in thickness-gradient Hf0.5Zr0.5O2 libraries
Y. Liu, Y.-X. Chen, H. Liang, I. Takeuchi, S. V. Kalinin. arXiv:2602.24216 (2026).
Real-time experiment–theory closed-loop interaction for autonomous materials science
H. Liang, C. Wang, D. Kirsch, R. Pant, A. McDannald, A. G. Kusne, J. Zhao, I. Takeuchi. Sci. Adv. 11, eadu7426 (2025).
Demonstration of real-time autonomous metal additive manufacturing
H. Liang, H. Huang, C. Sanjurjo-Rodriguez, N. Young, P. Kabirifar, A. G. Kusne, Y. Lee, I. Takeuchi, J. Zhao, S. S. Babu. ChemRxiv preprint, doi:10.26434/chemrxiv.15001156/v1.
Application of machine learning to RHEED images for automated structural phase mapping
H. Liang, V. Stanev, A. G. Kusne, Y. Tsukahara, K. Ito, R. Takahashi, M. Lippmaa, I. Takeuchi. Phys. Rev. Materials 6, 063805 (2022).
Benchmarking active learning strategies for materials optimization and discovery
A. Wang, H. Liang, A. McDannald, I. Takeuchi, A. G. Kusne. Oxford Open Materials Science 2(1) (2022).
A low-cost robot science kit for education with symbolic regression for hypothesis discovery
L. Saar, H. Liang, A. Wang, A. McDannald, E. Rodriguez, I. Takeuchi, A. G. Kusne. MRS Bulletin 47, 881–885 (2022).
A semi-supervised deep-learning approach for automatic crystal structure classification
S. Lolla, H. Liang, A. G. Kusne, I. Takeuchi, W. Ratcliff. J. Appl. Crystallogr. 55(4), 882–889 (2022).
CRYSPNet: Crystal structure predictions via neural network
H. Liang, V. Stanev, A. G. Kusne, I. Takeuchi. Phys. Rev. Materials 4, 123802 (2020).

Languages

English / Mandarin / Cantonese
Haotong Liang · Résumé 2 / 2