CV

Research Assistant

IBM-Illinois Discovery Accelerator Institute (IIDAI)

• With Prof. Han Zhao and Dr. Hendrick Hamann.
• Designed flexible and efficient foundation models for multimodal and hyperspectral geospatial data.
• Researched foundation models for spatiotemporal and multi-source geospatial data.
• Developed large-scale geospatial datasets and pipelines for multi-source acquisition, spatial and temporal indexing, alignment, metadata management, quality control, and scalable preprocessing.
• Worked on a large-scale event-centric InSAR dataset, including event discovery, geospatial sampling, temporal organization, preprocessing, and quality assessment.

Applied Scientist Intern

Amazon — Customer Trust

• With Prof. Rui Song, Prof. Hengrui Cai, and Prof. Sheng Wang.
• Researched conditioned text-embedding synthesis via an autoregressive conditional variational autoencoder.

Applied Scientist Intern

Amazon — Merch on Demand

• With Dr. Zhixiong Zhang.
• Enhanced text controllability and identity fidelity for personalized image generative models.

Research Intern

Shanghai AI Lab

• With Prof. Bin Zhao.
• Researched fully self-supervised depth from defocus.

Ph.D.

University of Illinois Urbana-Champaign

Electrical and Computer Engineering

• Expected graduation: May 2027. GPA: 4.0.
• Advisor: Prof. Han Zhao.

Bachelor of Science

University of Illinois Urbana-Champaign

Computer Engineering

• GPA: 3.97.
• Highest Honors at Graduation.

LESSViT: Robust Hyperspectral Representation Learning under Spectral Configuration Shift

Advances in Neural Information Processing Systems

Haozhe Si, Y. Wan, Y. Wang, M. Do, Han Zhao

Moment Alignment: Unifying Gradient and Hessian Matching for Domain Generalization

Conference on Uncertainty in Artificial Intelligence

Y. Chen, Haozhe Si, G. Zhang, Han Zhao

Towards a Foundation Model for Multi-modal and Multi-spectral Geospatial Data

TerraBytes Workshop at ICML

Haozhe Si, Y. Wan, M. Do, V. Deepak, Han Zhao, H. F. Hamann

Furry IP-Adapter: Preserving the Text-to-Image Capability of Stable Diffusion in IP-Adapter

Amazon Computer Vision Conference Workshop on Multimodal Generation

Haozhe Si, X. Zhang, Z. Zhang

Enhancing Compositional Generalization via Compositional Feature Alignment

Transactions on Machine Learning Research

Haoxiang Wang, Haozhe Si, Huajie Shao, Han Zhao

Fully Self-Supervised Depth Estimation from Defocus Clue

IEEE/CVF Conference on Computer Vision and Pattern Recognition

Haozhe Si, Bin Zhao, Dong Wang, Mulin Chen, Yunpeng Gao, Zhigang Wang, Xuelong Li

Provable Domain Generalization via Invariant-Feature Subspace Recovery

International Conference on Machine Learning

Haoxiang Wang, Haozhe Si, Bo Li, Han Zhao