Nikhila Ravi

Career

Nikhila Ravi's earlier Meta AI profile describes Ravi as a research engineer at Facebook AI Research focused on computer vision and 3D understanding, after earlier work at Facebook on tools and platforms for machine learning training and deployment for internal teams. Ravi is now a Director at Meta Superintelligence Labs and co-leads the Segment Anything (SAM) team.⁶¹

Research

Ravi is first author of the 2020 PyTorch3D paper, describing a library for 3D deep learning, and co-authored 3D vision work including Worldsheet (ICCV 2021), Omnivore (CVPR 2022) and the Omni3D benchmark for 3D object detection (CVPR 2023).⁷⁸

Ravi co-led the development of Segment Anything (2023), SAM 2 (2024), which extended promptable segmentation from images to video, and SAM 3 (2025), which segments objects from concept prompts and lists Ravi as a project lead. Ravi also co-authored the 2025 Perception Encoder and PerceptionLM papers presented at NeurIPS 2025.¹³²⁹⁴⁸

Open-source work

The SAM 2 repository releases model checkpoints, demo code and training code under the Apache 2.0 license, and Ravi's site notes the later release of SAM 2.1 checkpoints with a developer suite for fine-tuning and the web demo. SAM 3 was released with a SAM Playground, followed by SAM 3.1, which Ravi's site describes as doubling video processing speed.⁵¹

Work and contributions

• Segment Anything (SAM, SAM 2 and SAM 3) Open weights

  First author of the SAM 2 paper (2024), co-author of the original Segment Anything paper (2023), and one of the project leads named in the SAM 3 repository README. The SAM 2 repository releases checkpoints, demo code, and training code under Apache 2.0.²³⁵⁴

• SAM 2.1 Hiera-Large Open weights

  First author of the SAM 2 paper, which the SAM 2.1 Hiera-Large model card cites.²

• SAM 3.1 Open weights

  Co-author of the SAM 3 paper, which the facebookresearch/sam3 repository cites; that repository also hosts the code for the SAM 3.1 checkpoints.⁹⁴

Timeline

1. Jul 16, 2020Posted "Accelerating 3D Deep Learning with PyTorch3D" as first author.⁷
2. Apr 5, 2023Co-authored the Segment Anything paper, later an ICCV 2023 oral.³⁸
3. Aug 1, 2024Posted "SAM 2: Segment Anything in Images and Videos" as first author.²
4. Nov 20, 2025Posted "SAM 3: Segment Anything with Concepts" as one of the project leads.⁹⁴
5. 2026Had the SAM 3 paper, written with co-authors at Meta, accepted to ICLR 2026.¹⁸

Other documented roles

Roles the sources document that are not confirmed as current.

• Research Engineer, FAIR (earlier Meta AI profile), Meta (Facebook AI Research)⁶

Honors and recognition

• ICCV 2023 Best Paper Honourable Mention (Segment Anything) · International Conference on Computer Vision, 2023⁸
• ICLR 2025 Best Paper Honourable Mention (SAM 2) · International Conference on Learning Representations, 2025⁸¹
• Young Engineer of the Year Award · Royal Academy of Engineering¹

Where they work

• Meta 106 people in the wiki

Related people

• Matt DeitkeAlso at Meta
• Alexander KirillovAlso at Meta
• Xinlei ChenAlso at Meta
• Saining XieAlso at Meta
• Christoph FeichtenhoferAlso at Meta
• Jyoti AnejaMultimodal models · Computer vision
• Shusheng YangMultimodal models · Computer vision
• Hassan ShapourianMultimodal models · Computer vision
• João CarreiraComputer vision · Multimodal models
• Fei-Fei LiComputer vision · Multimodal models