PeopleofSuperintelligence

Researcher

Weiyao Wang

Weiyao Wang is an AI researcher whose work spans nonconvex optimization, multimodal training, open-world vision and 3D reconstruction. He graduated from Duke with highest distinction in 2018 and earned a bachelor's degree in mathematics and computer science; he coauthored Stabilized SVRG in 2019. At FAIR Meta, he coauthored Gradient Blending in 2020, the UVO benchmark in 2021 and Generic Grouping Networks in 2022. His later work includes the EgoTracks dataset, ICON for NeRF training from video and SAM 3D, which predicts an object's geometry, texture and layout from one image; the SAM 3D paper received a CVPR 2026 Best Paper Honorable Mention. He is a member of technical staff at Thinking Machines Lab, where he focuses on multimodal foundation models, 3D world modeling systems and large-scale perception architectures.

Badges

Timeline

  1. 2018–2019

    Duke graduation and SVRG

    In 2018, he graduates from Duke with highest distinction. Before entering industry, he earns a bachelor's degree in mathematics and computer science, with Rong Ge and Robert Calderbank as his advisers. In 2019, he coauthors Stabilized SVRG with Ge, Zhize Li and Xiang Wang, a simple SVRG variant that finds approximate second-order stationary points in nonconvex optimization.

  2. 2020

    Gradient Blending at CVPR

    At FAIR Meta, he collaborates with and receives mentorship from Du Tran. In 2020, he coauthors “What Makes Training Multi-Modal Classification Networks Hard?” with Tran and Matt Feiszli at CVPR; the paper introduces Gradient Blending, which weights modalities by their overfitting behavior.

  3. 2021

    The UVO benchmark

    In 2021, he coauthors UVO, a benchmark for open-world class-agnostic object segmentation in videos. He also co-organizes the Open-World Video Object Detection and Segmentation Challenge at ICCV.

  4. 2022

    Generic Grouping Networks

    In 2022, he coauthors the CVPR paper “Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity.” It proposes Generic Grouping Networks, mask proposals built without semantic supervision.

  5. 2023–2024

    EgoTracks and ICON

    In 2023, he coauthors EgoTracks, a long-term egocentric visual object tracking dataset sourced from Ego4D. In 2024, he coauthors ICON, a CVPR optimization procedure that trains NeRFs from 2D video frames without pose initialization.

  6. November 2025

    SAM 3D from one image

    In November 2025, SAM 3D launches and Wang is among its authors. The generative model predicts an object's geometry, texture and layout from a single image.

  7. 2026

    A CVPR honorable mention

    In 2026, SAM 3D receives a Best Paper Honorable Mention at CVPR, one of two papers named for that honor by the Awards Committee.

  8. Since April 2026

    Member of technical staff at TML

    Before joining TML, he works as an AI researcher and tech lead at Meta Superintelligence Lab. After eight years at Meta—his first job after college—he joins Thinking Machines Lab in April 2026 as a member of technical staff. His research there focuses on multimodal foundation models, 3D world modeling systems and large-scale perception architectures.

Career

  1. Since 2026Thinking Machines LabMember of technical staff

Education

  1. 2018Duke UniversityBachelor’s degree

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Weiyao Wang. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  2. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Thinking Machines Lab. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  3. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Thinking Machines Lab. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.
  4. 5,000 frontier users would be 38.2% of the core labs' 13,100 reported staffIncludes Thinking Machines Lab. 5,000 frontier users would be 38.2% of the core labs' 13,100 reported staff. We test the estimate against 3,913 named model contributors and dated all-staff records across Google DeepMind, OpenAI and Anthropic, plus six additional frontier-grade labs.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Weiyao Wang yet. Send us a free-licensed photo.

  1. Weiyao Wang sites.google.com. Checked 21 October 2026.
  2. Undergraduate Awards Presented 2018 cs.duke.edu. Checked 21 October 2026.
  3. Stabilized SVRG: Simple Variance Reduction for Nonconvex Optimization arxiv.org. Checked 21 October 2026.
  4. What Makes Training Multi-Modal Classification Networks Hard? (CVPR 2020 Open Access Repository) openaccess.thecvf.com. Checked 21 October 2026.
  5. Unidentified Video Objects: A Benchmark for Dense, Open-World Segmentation (ICCV 2021 Open Access Repository) openaccess.thecvf.com. Checked 21 October 2026.
  6. Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity (CVPR 2022 Open Access Repository) openaccess.thecvf.com. Checked 21 October 2026.
  7. EgoTracks: A Long-term Egocentric Visual Object Tracking Dataset arxiv.org. Checked 21 October 2026.
  8. ICON: Incremental CONfidence for Joint Pose and Radiance Field Optimization (CVPR 2024 Open Access Repository) openaccess.thecvf.com. Checked 21 October 2026.
  9. SAM 3D: 3Dfy Anything in Images arxiv.org. Checked 21 October 2026.
  10. CVPR 2026 Honors the Year's Most Innovative Computer Vision and AI Research cvpr.thecvf.com. Checked 21 October 2026.
  11. Weiyao Wang | alphaXiv alphaxiv.org. Checked 21 October 2026.
  12. Meta's loss is Thinking Machines' gain techcrunch.com. Checked 21 October 2026.