PeopleofSuperintelligence

Researcher

Romain Beaumont

Also known as rom1504

Romain Beaumont is a machine learning engineer and a founder of LAION, which released the open image-text datasets LAION-400M and LAION-5B. After computer science studies, he was a PhD student at Université Paris-Sud from 2014 to 2017, then worked at Criteo from 2017 to 2022 and at Google from 2022 to 2025. He coauthored LAION-400M in 2021, then built and ran the dataset preparation pipeline for LAION-5B in 2022 and was second author of the paper, which was named an Outstanding Paper at NeurIPS 2022. In 2022 he also trained the large ViT-H/14 and ViT-g/14 OpenCLIP models; in 2023 he was second author of a CVPR study of CLIP scaling laws, and LAION won the Falling Walls Science Breakthrough of the Year award, which names him among its founding members. Since 2025 he is a Staff Research Engineer at Google DeepMind.

Badges

Timeline

  1. 2011–2017

    Computer science and PhD studies

    He studies computer science and applied mathematics at the École Nationale Supérieure d'Informatique pour l'Industrie et l'Entreprise from 2011 to 2014. From 2014 to 2017 he is a PhD student at Université Paris-Sud and the Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur.

  2. 2017–2022

    Machine learning at Criteo

    From 2017 to 2022 he works at Criteo in France, starting as a software engineer and reaching Staff Machine Learning Engineer in 2021.

  3. 2021

    The LAION-400M dataset

    He is a LAION founder, and in 2021 he coauthors LAION-400M, an open dataset of 400 million CLIP-filtered image-text pairs released with CLIP embeddings and kNN indices for similarity search. The short paper is accepted at the Data Centric AI NeurIPS Workshop 2021.

  4. 2022

    LAION-5B and large OpenCLIP models

    In March 2022 he announces LAION-5B, a dataset of 5.85 billion CLIP-filtered image-text pairs, and builds and runs its dataset preparation pipeline with img2dataset, CLIP inference and autofaiss. He also develops clip-retrieval, which embeds LAION images and captions with CLIP for k-NN search. In September 2022 he trains the ViT-H/14 and ViT-g/14 OpenCLIP models on the Stability cluster, and the H/14 model reaches 78.0% zero-shot top-1 accuracy on ImageNet.

  5. 2022

    A NeurIPS Outstanding Paper

    He is the second author, after Christoph Schuhmann, of the LAION-5B paper, which is named an Outstanding Paper at NeurIPS 2022 and appears in the conference's Datasets and Benchmarks track.

  6. 2022–2025

    Software engineer at Google

    He joins Google in 2022 as a software engineer and becomes a staff software engineer in 2024.

  7. 2023

    Scaling laws and Falling Walls

    In 2023 he is second author of "Reproducible scaling laws for contrastive language-image learning", a CVPR 2023 study of CLIP models trained on up to two billion image-text pairs from the public LAION dataset. LAION wins the Falling Walls Science Breakthrough of the Year 2023 award in the Science & Innovation Management category, and he is named among its founding members.

  8. Since 2025

    Staff Research Engineer at DeepMind

    Since 2025 he is a Staff Research Engineer at Google DeepMind. In August 2026 he is among the authors of LAION-BVD, an open video dataset built from 1.3 billion video URLs collected from CommonCrawl, with 80 million downloaded videos totaling 10 million hours.

Career

  1. Since 2025Google DeepMindStaff Research Engineer
  2. NowLAIONFounder
  3. 2022–2025GoogleSoftware Engineer, Staff Software Engineer
  4. 2017–2022CriteoSoftware Engineer, Staff Machine Learning Engineer

Education

  1. 2017Université Paris-SudPhD
  2. 2017Laboratory of Computer Science for Mechanics and Engineering SciencesPhD
  3. 2014National School of Computer Science for Industry and Business

From the biography.

  1. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  2. Google has the highest combined h-index among AI organizationsIncludes Google. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  3. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.
  4. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Google. 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.

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 Romain Beaumont yet. Send us a free-licensed photo.

  1. Romain Beaumont (Google DeepMind): computer vision — Lacuna lacuna.tiptreesystems.com. Checked 22 October 2026.
  2. [2111.02114] LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairs arxiv.org. Checked 22 October 2026.
  3. Team | LAION laion.ai. Checked 22 October 2026.
  4. LAION-5B: A NEW ERA OF OPEN LARGE-SCALE MULTI-MODAL DATASETS | LAION laion.ai. Checked 22 October 2026.
  5. Automating image collection – Replicate blog replicate.ai. Checked 22 October 2026.
  6. Large scale openCLIP: L/14, H/14 and g/14 trained on LAION-2B | LAION laion.ai. Checked 22 October 2026.
  7. NeurIPS 2022 Awards neurips.cc. Checked 22 October 2026.
  8. [2210.08402] LAION-5B: An open large-scale dataset for training next generation image-text models arxiv.org. Checked 22 October 2026.
  9. [2212.07143] Reproducible scaling laws for contrastive language-image learning arxiv.org. Checked 22 October 2026.
  10. Christoph Schuhmann, Richard Vencu, Romain Beaumont, Robert Kaczmarczyk & Jenia Jitsev | Falling Walls falling-walls.com. Checked 22 October 2026.
  11. [2608.24845] LAION-BVD: A 10-Million-Hour Open Video Dataset for Multimodal Pre-training arxiv.org. Checked 22 October 2026.