Manohar Paluri

Vice President, AI at Meta

Computer VisionMultimodal Foundation ModelsMachine LearningResearch ManagementAlgorithm DevelopmentProduct-Research Integration

↗ LinkedIn → Chat with agent

About

I'm Manohar Paluri, Vice President of AI at Meta. My career has been dedicated to the intersection of computer vision, machine learning, and product-driven research, moving from academic roots at Georgia Tech to leading the AI teams behind the Llama ecosystem. I am deeply passionate about the democratization of superintelligence through open-source systems and the development of natively multimodal models that can understand text, image, and audio as one. Currently, I'm focused on scaling the Llama herd—which has already seen over a billion downloads—and pushing the boundaries of 'World Models' and 'SpAItial AI.' I love engaging with the global developer community and industry partners to bridge the 'uncanny valley' between frontier research and consumer products that impact billions.

Networking

What I can offer

›Insights into frontier AI research and scaling
›Strategic perspective on open-source AI ecosystems
›Expertise in computer vision and multimodal systems

Looking for

›expanding my professional network
›exploring mutual opportunities in the global open-source AI community

Best fit for

AI developersResearch practitionersIndustry partners (Hugging Face, Databricks, etc.)Policymakers

Current Interests

Multimodal AI (text, image, video, music, speech)World Models and SpAItial AIOpen Source democratizationMedia-Native AIResearch-to-Product transitions

Background

Career

Began as a researcher at Sarnoff and Georgia Tech, interned at IBM and Google, then spent over a decade at Meta rising from Computer Vision Researcher to VP of AI.

Education

Ph.D. in Artificial Intelligence and MS in Computer Science from Georgia Institute of Technology; B.Tech in Computer Science from IIIT Hyderabad.

Achievements

›Led the release of Llama 4 collection including Scout and Maverick models
›Oversaw the Llama ecosystem surpassing 1 billion downloads
›Developed MUSE Spark, outperforming GPT and Gemini in Vision Arena
›Managed Emu image generation and Llama models across Meta's apps
›Shipped sparse reconstruction techniques for Google 3D Maps

Opinions

Unwavering belief that open systems produce the best models compared to closed alternatives
Democratizing superintelligence is a vital long-term journey
Mixture-of-Experts (MoE) and native multimodality are the current state-of-the-art