Research interests: I am broadly interested designing physics motivated machine learning algorithms for solving real-world problems related to material design, drug design, and other engineering domains. I study how domain-specific inductive biases can be incorporated into neural networks to improve generalizability and interpretability. I work closely with chemists, material scientists, and other domain experts to deploy machine learning algorithms to accelerate existing workflows and gain scientific insights.
Previously: I got my PhD in Materials Science and Engineering at MIT in 2020, advised by Jeffrey C. Grossman. I got my BS in Chemistry at Peking University in 2015. I was a research intern at DeepMind, working with Pushmeet Kohli and James Kirkpatrick in 2019/05. I was a research intern at Google X in 2019/02.
recent news [earlier]
|2021/12||I will give a contributed talk (top 2 %) about our latest CDVAE paper at the NeurIPS ML4Physics workshop on Dec. 13, 2021.|
|2021/10||Excited to share our latest preprint “Crystal Diffusion Variational Autoencoder for Periodic Material Generation” on arXiv. Update (2021/12): code and data are now available on our github repo.|
|2021/05||Our lastest work “Charting lattice thermal conductivity for inorganic crystals and discovering rare earth chalcogenides for thermoelectrics is published on Energy & Environmental Science.|
selected publications [full list]
- Crystal Diffusion Variational Autoencoder for Periodic Material GenerationarXiv preprint arXiv:2110.06197 2021
- Accelerating the screening of amorphous polymer electrolytes by learning to reduce random and systematic errors in molecular dynamics simulationsNeurIPS ML4Molecules workshop 2020Spotlight Talk, full paper at arXiv
- Graph dynamical networks for unsupervised learning of atomic scale dynamics in materialsNature communications 2019
- Machine learning enabled computational screening of inorganic solid electrolytes for suppression of dendrite formation in lithium metal anodesACS central science 2018
- Crystal graph convolutional neural networks for an accurate and interpretable prediction of material propertiesPhysical review letters 2018