PeopleofSuperintelligence

Researcher

Xiang Zhang

Xiang Zhang is an AI researcher whose work spans convolutional vision, character-based language models, multilingual text classification, and image-text retrieval. He co-developed OverFeat in 2013, coauthored character-level text models at NYU, and compared language encodings across four languages. He contributed to Google Research's DeepCTRL and Pic2Word, and wrote about serving Llama-2 and using generative AI in consumer products. As of October 2026, he is Chief AI Scientist at China Premium Smart Appliances Innovation Center.

Now at
China Premium Smart Appliances Innovation Center
Works
53

Publication numbers from OpenAlex, as of 22 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. 2013

    OverFeat wins at ImageNet

    With Pierre Sermanet, David Eigen, Michael Mathieu, Rob Fergus and Yann LeCun, he coauthors OverFeat in 2013, an integrated convolutional-network framework for classification, localization and detection. The system wins the 2013 ImageNet localization task.

  2. 2015

    Character-level text classification

    At NYU, he and Junbo Zhao coauthor Character-level Convolutional Networks for Text Classification in 2015. The paper applies ConvNets directly to characters across large datasets and reports state-of-the-art or competitive results.

  3. 2017

    Multilingual text encodings

    He and Yann LeCun compare 473 encoding models across 14 text-classification datasets in Chinese, English, Japanese and Korean. Their 2017 study finds that fastText with character-level n-grams gives the best results.

  4. 2021

    DeepCTRL rule control

    At NeurIPS 2021, he contributes to DeepCTRL, a method that adds rule representations to neural networks and lets users control rule strength at inference without retraining.

  5. 2023

    Pic2Word image retrieval

    He coauthors Pic2Word in 2023, a zero-shot composed-image retrieval model trained on weakly labeled image-caption pairs and unlabeled images.

  6. September 2023

    Serving Llama-2 7B

    In September 2023, he publishes a guide to serving Llama-2 7B with llama.cpp, NVIDIA CUDA and Ubuntu 22.04.

  7. January 2024

    Generative AI in products

    In January 2024, he argues that generative AI can add conversational and drawing features to consumer products. His example refrigerator can order groceries and answer general questions.

  8. As of October 2026

    Chief AI Scientist

    As of October 2026, he is Chief AI Scientist at China Premium Smart Appliances Innovation Center. His PhD advisor was Yann LeCun, and he has long worked on foundational AI research and frontier exploration.

  1. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Southeast University. We estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.
  2. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Southeast University. 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.
  3. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Southeast University. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  4. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.

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 Xiang Zhang yet. Send us a free-licensed photo.

  1. OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks arxiv.org. Checked 20 October 2026.
  2. Character-level Convolutional Networks for Text Classification arxiv.org. Checked 20 October 2026.
  3. NIPS 2015 List of Accepted Papers nips.cc. Checked 20 October 2026.
  4. Which Encoding is the Best for Text Classification in Chinese, English, Japanese and Korean? arxiv.org. Checked 20 October 2026.
  5. Controlling Neural Networks with Rule Representations research.google. Checked 20 October 2026.
  6. Pic2Word: Mapping Pictures to Words for Zero-shot Composed Image Retrieval research.google. Checked 20 October 2026.
  7. Serving Llama-2 7B using llama.cpp with NVIDIA CUDA on Ubuntu 22.04 xzh.me. Checked 20 October 2026.
  8. Thoughts on AIGC for Non-AI Industries xzh.me. Checked 20 October 2026.
  9. Models Will Get Cheaper, Problems Will Get More Valuable: From Today’s AI Boundaries to the Problem-Solving Industry 20 Years From Now xzh.me. Checked 20 October 2026.