Ramesh Raskar is an associate professor at the MIT Media Lab, founder and director of its Camera Culture research group, and principal investigator of Project NANDA, which is developing open infrastructure for autonomous AI agents. His inventions range from cameras that reconstruct scenes around corners to protocols that help independently operated software agents discover and trust one another.
Raskar studied engineering at the College of Engineering, Pune, and completed doctoral research at the University of North Carolina at Chapel Hill. At Mitsubishi Electric Research Laboratories, he developed projector-based augmented reality, multiflash photography, computational approaches to motion blur, and immersive displays.
He joined MIT and founded Camera Culture in 2008. There, he pioneered femto-photography, ultrafast imaging that captures light in motion and uses computational reconstruction to reveal objects hidden around corners. The work contributed to his 2016 Lemelson-MIT Prize.
Raskar also co-founded EyeNetra, which developed smartphone-based vision assessments, and led Safe Paths, an open-source, privacy-conscious COVID-19 contact-tracing initiative. His group advanced privacy-preserving split learning, enabling organizations to collaborate on machine-learning systems without sharing raw data. He also led innovation efforts at Facebook spanning digital health, connectivity, imaging, and virtual and augmented reality. In 2024, he was elected a National Academy of Inventors Fellow.
Building an open internet for AI agents
Through Project NANDA—Networked AI Agents in a Decentralized Architecture—Raskar is building an open agentic web where software agents can locate counterparts, verify credentials, exchange information, and coordinate across organizational boundaries.
His central concern is that agents locked inside proprietary platforms cannot participate in a wider digital economy. Unlike websites, autonomous agents initiate tasks, delegate work, move between hosts, and negotiate with other agents; their infrastructure must therefore handle changing endpoints, portable identities, authorization, and trust.
Three priorities shape that architecture:
- The NANDA Index and verified AgentFacts: A proposed agent-discovery architecture links identities to signed records describing capabilities, credentials, permissions, and reachable endpoints, with adaptive routing and privacy-sensitive access.
- Trust without a central gatekeeper: Raskar’s enterprise-focused research addresses impersonation, capability spoofing, sensitive-data leakage, and interoperability across competing agent protocols.
- Discovery, commerce, and decentralized coordination: His vision extends from identifying trustworthy agents to enabling transactions and collective action. NandaTown provides a simulation environment for testing negotiation, auctions, messaging, consensus, and other interactions before agent networks become operationally critical.
From optical sensing to public-health technology and agent networks, Raskar develops infrastructure that makes difficult-to-observe systems legible while preserving distributed control.
