PeopleofSuperintelligence

Researcher

Zi-Yi Dou

Zi-Yi Dou is a computer scientist whose research connects language, vision and AI agents. They earned a computer science B.S. with a statistics minor at Nanjing University in 2018 and an M.S. in Language Technologies at Carnegie Mellon University in 2020, then completed a computer science PhD at UCLA under Nanyun (Violet) Peng in 2025; they received an Amazon PhD Fellowship in 2022. Their research moved from document sentiment analysis and neural machine translation to FIBER, a vision-language model, and self-collected navigation and language-agent training methods. They coauthored Ctrl-R, which guides model rollouts to explore diverse reasoning patterns. After research roles at Apple AI/ML and xAI, they are a Research Scientist at Meta MSL, where they work on multimodal and computer-use agents for Muse Spark.

Badges

Timeline

  1. 2014–2020

    Computing and language degrees

    At Nanjing University, they earn a B.S. in computer science with a statistics minor in 2018. They complete an M.S. in Language Technologies at Carnegie Mellon University under Graham Neubig in 2020.

  2. 2017–2019

    Sentiment and translation papers

    In 2017, they publish a deep memory network for document-level sentiment analysis that combines user and product information. In 2018, they coauthor a neural machine translation paper that uses layer aggregation and multi-layer attention to expose signals across encoder and decoder layers. Their 2019 Dynamic Layer Aggregation paper applies routing-by-agreement to combine model layers dynamically.

  3. 2020–2025

    UCLA PhD and Amazon fellowship

    They complete a computer science PhD at UCLA in January 2025 under Nanyun (Violet) Peng. In 2022, they receive the Amazon PhD Fellowship. Their research connects language to perception and action through vision-language pretraining and embodied agents.

  4. 2022

    FIBER vision-language architecture

    At NeurIPS 2022, they coauthor FIBER, a vision-language model that pushes multimodal fusion into image and text backbones through cross-attention and uses a two-stage pretraining strategy.

  5. 2023–2024

    Navigation and language-agent papers

    In 2023, they coauthor Masked Path Modeling, which pretrains navigation agents on self-collected paths recorded as they explore environments. At EMNLP 2024, they coauthor Re-ReST, which uses a reflector and external-environment feedback to refine generated examples for language-agent training.

  6. January–November 2025

    Research scientist at Apple

    From January to November 2025, they work as a Research Scientist at Apple AI/ML on multimodal post-training and agents for Apple Foundation Models.

  7. November 2025–March 2026

    Member of Technical Staff at xAI

    From November 2025 to March 2026, they work as a Member of Technical Staff at xAI on mid-training and post-training for computer-use agents.

  8. Since April 2026

    Research scientist at Meta MSL

    In July 2026, they coauthor Ctrl-R, which guides rollouts to explore diverse reasoning patterns, based on work completed at Apple. Since April 2026, they are a Research Scientist at Meta MSL, where they work on multimodal and computer-use agents for Muse Spark.

Career

  1. Since 2026Meta Superintelligence LabsResearch Scientist
  2. 2025AppleResearch Scientist
  3. 2025–2026xAIMember of Technical Staff

Education

  1. 2025University of California, Los AngelesPhD
  2. 2020Carnegie Mellon UniversityM.S.
  3. 2018Nanjing UniversityB.S.

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Zi-Yi Dou. 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. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Zi-Yi Dou. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  3. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Meta. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  4. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Meta Superintelligence Labs. 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.

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 Zi-Yi Dou yet. Send us a free-licensed photo.

  1. Zi-Yi Dou zdou0830.github.io. Checked 21 October 2026.
  2. Capturing User and Product Information for Document Level Sentiment Analysis with Deep Memory Network aclanthology.org. Checked 21 October 2026.
  3. Exploiting Deep Representations for Neural Machine Translation aclanthology.org. Checked 21 October 2026.
  4. Dynamic Layer Aggregation for Neural Machine Translation with Routing-by-Agreement ojs.aaai.org. Checked 21 October 2026.
  5. 2022 Amazon Fellows | Science Hub for Humanity and Artificial Intelligence sciencehub.ucla.edu. Checked 21 October 2026.
  6. Coarse-to-Fine Vision-Language Pre-training with Fusion in the Backbone proceedings.neurips.cc. Checked 21 October 2026.
  7. Masked Path Modeling for Vision-and-Language Navigation aclanthology.org. Checked 21 October 2026.
  8. Re-ReST: Reflection-Reinforced Self-Training for Language Agents aclanthology.org. Checked 21 October 2026.
  9. Learning Structured Reasoning via Tractable Trajectory Control - Apple Machine Learning Research machinelearning.apple.com. Checked 21 October 2026.