avatar

Theodore Zhao

Zhengde Zhao 赵正德

Senior Applied Scientist
Microsoft
Redmond, WA, USA

About Me

I am a Researcher at Microsoft focusing on designing and training vision and multimodal foundation models. I received my PhD and Master degrees in Applied Mathematics from University of Washington with research focus on probabilistic and multiscale temporal modelling, advised by Prof. Tim Leung. During PhD, I interned at Microsoft Research, Harborview Medical Center, Rotella Capital, and Parametric. I received my Bachelor degree in Physics from Fudan University in Shanghai, China.

My research interest broadly covers:
  • Visoin Modelling: Self-supervised Learning, Unified Representation, World Models.
  • Multimodal Modelling: Multimodal Alignment, Vision-Language Models, Pixel Grounding.
  • General Machine Learning: Reinforcement Learning, Pareto Optimization, Multi-objective Learning.

I am particularly interested in learning generic intelligence from the vision format (images, videos) and welcome any discussion/collaboration!

News

  • [May 2026] STELLAR is accepted by ICML 2026. See you in Seoul!
  • [Feb. 2026] We released STELLAR, a unified vision representation learning framework supporting both rich semantics and high-quality reconstruction from only 16 tokens.
  • [Jun. 2025] The extension of BiomedParse to 3D images won the first place in CVPR 2025: Foundation Models for Text-guided 3D Biomedical Image Segmentation Challenge!
  • [Mar. 2025] BoltzFormer is accepted at CVPR 2025! See you in Nashville!
  • [Nov. 2024] BiomedParse is in Nature Methods! We are also releasing the code, model, and a million scale text promptable medical image segmentation dataset.
  • [May. 2024] Proud to lead the BiomedPrase project. We presented a foundation model for joint segmentation, detection and recognition of biomedical objects across nine modalities. Various medical image analysis tasks are unified with natural language prompting.
  • [May. 2024] Our paper on Error Detection is accepted by ACL 2024. See you in Bangkok!
  • [April. 2024] I am giving a talk at 2024 Joint Statistical Meetings on 08/05/2024 in Portland.

Recent Works

  1. Publication Teaser Image
    Learning Sparse Visual Representations via Spatial-Semantic Factorization
    Theodore Zhengde Zhao, Sid Kiblawi, Jianwei Yang, Naoto Usuyama, Reuben Tan, Noel C. Codella, Tristan Naumann, Hoifung Poon, Mu Wei
    ICML, 2026
    [PAPER] [CODE]
  2. Publication Teaser Image
    Masked-Diffusion Autoencoders for 3D Medical Vision Representation Learning
    Jiachen Tu, Guanghui Qin, Theodore Zhengde Zhao, Jeya Maria Jose Valanarasu, Sheng Zhang, Tristan Naumann, Fan Lam, Sheng Wang, Hoifung Poon
    CVPR, 2026
  3. Publication Teaser Image
    Multimodal Reinforcement Learning with Agentic Verifier for AI Agents
    Reuben Tan, Baolin Peng, Zhengyuan Yang, Hao Cheng, Oier Mees, Theodore Zhao, Andrea Tupini, Isar Meijier, Qianhui Wu, Yuncong Yang, Lars Liden, Yu Gu, Sheng Zhang, Xiaodong Liu, Lijuan Wang, Marc Pollefeys, Yong Jae Lee, Jianfeng Gao
    arXiv Preprint, 2025
    [PAPER]
  4. Publication Teaser Image
    BiomedParse-V: Scaling Foundation Model for Universal Text-guided Volumetric Biomedical Image Segmentation
    Theodore Zhao, Ho Hin Lee, Alberto Santamaria-Pang, Noel C. Codella, Sid Kiblawi, Yu Gu, Yu Fang, Wen Xuan Teng, Naiteek Sangani, Ivan Tarapov, Matthew P. Lungren, Matthias Blondeel, Tristan Naumann, Naoto Usuyama, Sheng Wang, Paul Vozila, Hoifung Poon, Mu Wei^
    Accepted to CVPR 2025 MedSegFM, 2025
    [PAPER] [CODE]
  5. Publication Teaser Image
    Boltzmann Attention Sampling for Image Analysis with Small Objects
    Theodore Zhao*, Sid Kiblawi*, Naoto Usuyama, Hohin Lee, J Samuel Preston, Hoifung Poon, Mu Wei^
    CVPR, 2025
    [PAPER] [CODE]
  6. Publication Teaser Image
    A foundation model for joint segmentation, detection and recognition of biomedical objects across nine modalities
    Theodore Zhao*, Yu Gu*, Jianwei Yang, Naoto Usuyama, Ho Hin Lee, Sid Kiblawi, Tristan Naumann, Jianfeng Gao, Angela Crabtree, Jacob Abel, Christine Moung-Wen, Brian Piening, Carlo Bifulco, Mu Wei^, Hoifung Poon^, Sheng Wang^
    Nature Methods, 2024
    [PAPER] [DEMO] [BLOG] [CODE] [MODEL] [DATA]
  7. Publication Teaser Image
    Pareto Optimal Learning for Estimating Large Language Model Errors
    Theodore Zhao, Mu Wei, J Samuel Preston, Hoifung Poon
    ACL, 2024
    [PDF]

Some Previous Works

  1. Publication Teaser Image
    A Noisy Fractional Brownian Motion Model for Multiscale Correlation Analysis of High-Frequency Prices
    Tim Leung*, Theodore Zhao*
    Mathematics, 2024
  2. Publication Teaser Image
    Fee for home delivery and monitoring of antiretroviral therapy for HIV infection compared with standard clinic-based services in South Africa: a randomised controlled trial
    Ruanne V Barnabas, Adam A Szpiro, Xolani Ntinga, Melissa Latigo Mugambi, Heidi van Rooyen, Andrew Bruce, Philip Joseph, Thulani Ngubane, Meighan L Krows, Torin T Schaafsma, Theodore Zhao, Frank Tanser, Jared M Baeten, Connie Celum, Alastair van Heerden, Siyabonga Nkala
    The Lancet HIV, 2022
    We conducted a randomised trial, the Deliver Health Study, of a fee for home delivery of ART compared d with free clinic ART delivery in South Africa.
  3. Publication Teaser Image
    Financial time series analysis and forecasting with Hilbert–Huang transform feature generation and machine learning
    Tim Leung*, Theodore Zhao*
    Applied Stochastic Models in Business and Industry, 2021
    [PDF]
    We present the method of complementary ensemble empirical mode decomposition (CEEMD) and Hilbert-Huang transform (HHT) for analyzing nonstation financial time series.

Misc

In my free time, I love stargazing. I also started making my own music recently.

Here are some of my music — Just a heads-up, you'll hear real SOUND when you hit “play”! 🎵 🎧

  • Entanglement

  • Pearl Waltz

  • À La Recherche Du Temps Perdu

  • Sunshine in the Winter

If my research/music interests you, send me an email! I'm always open to collaborations and discussions.🍺