Manohar Paluri

Manohar Paluri is a computer vision and multimodal AI researcher whose work runs from Facebook's early AI research group to the Llama models. He earned a B.Tech at IIIT Hyderabad in 2006, worked at Sarnoff Corporation, began a PhD at Georgia Tech, joined Facebook in 2012, and became a founding member of Facebook AI Research in 2013. He coauthored the C3D and R(2+1)D video papers, rose from manager to director to vice president at Meta, and counts Ego4D, Make-A-Video and the Llama models among his projects. IIIT Hyderabad gave him its Outstanding Alumni Service Award in 2025, and in July 2025 he became VP of Media Applied Research within Meta Superintelligence Labs.

Now at: Meta Superintelligence Labs
Badges: Founder
Bridge
Links: Website, Google Scholar, X, LinkedIn and GitHub

Timeline

2002–2012

Hyderabad, Sarnoff and Georgia Tech

He earns a B.Tech in computer science at IIIT Hyderabad in 2006 and joins Sarnoff Corporation as a member of technical staff. In 2009, he starts a computer science PhD at Georgia Tech, where he completes an MS and contributes to the GTSAM factor-graph library in Frank Dellaert's group.
2012–2015

A founding member of FAIR

In 2012, after a research internship at Google on 3D reconstruction for Google Maps, he moves to Facebook as a computer vision researcher and engineer. He is a founding member of Facebook AI Research (FAIR) in 2013.
2014–2015

The C3D video features

In 2014, he coauthors "Learning Spatiotemporal Features with 3D Convolutional Networks" with Du Tran, Lubomir Bourdev, Rob Fergus and Lorenzo Torresani. The paper learns C3D features with 3D convolutional networks trained on a large supervised video dataset, and a simple linear classifier on them outperforms state-of-the-art methods on 4 benchmarks. It appears at ICCV in 2015.
2015–2018

Automatic Alt Text and Lumos

From 2015, he is a manager and then a senior manager at Facebook. His projects in these years include Automatic Alt Text in 2016, which generates descriptions of images for screen readers, the Caffe2Go on-device mobile runtime in 2016, and the Lumos visual understanding platform in 2017.
2018–2021

Director and the R(2+1)D block

In July 2018, he becomes Director of AI at Facebook, leading applied computer vision and augmented reality teams that serve image and video understanding to hundreds of products across Meta's apps. His 2018 CVPR paper on R(2+1)D, coauthored with Du Tran, Heng Wang, Lorenzo Torresani, Jamie Ray and Yann LeCun, factorizes 3D convolutional filters into separate spatial and temporal components.
2021–2023

Ego4D and Make-A-Video

In July 2021, he becomes Senior Director of FAIR Accel, a group working on perception and language. His projects include Ego4D in 2021, a first-person video dataset of more than 3,000 hours with a benchmark suite, and in 2022 the text-and-sketch image generator Make-A-Scene and the text-to-video system Make-A-Video.
2023–2025

Leading Llama at Meta

In February 2023, he becomes Senior Director for Llama and generative AI at Meta, where Llama 2 is among his projects in 2023; in February 2024, he becomes Vice President. In 2024, he is among the authors of The Llama 3 Herd of Models, whose largest model is a dense Transformer with 405B parameters. IIIT Hyderabad gives him its Outstanding Alumni Service Award in 2025.
July 2025

Media Applied Research at MSL

In July 2025, he becomes VP of Media Applied Research within Meta Superintelligence Labs, where his group works on multimodal understanding and generation. Muse Spark, the first public release from the lab, is among his 2025 projects.