Panel · 12:00 · Models Scale

The Physical AI Stack

Minds in Motion at MIT, Physical AI Summit · 30 Oct 2026 · Cambridge, United States

About this session

From the compute that trains the models to the autonomy and networks that put them to work, what physical AI needs to scale, and where new kinds of computing come in.

On stage

moderatorDave BlundinFounder & General Partner, Link Ventures
panelistVijay GadepallySenior Staff, MIT Lincoln Lab
panelistSertac KaramanProfessor, MIT
panelistRamesh RaskarAssoc. Prof. MIT Media Lab
panelistMassimiliano VersaceGlobal Head of Emergent AI, Analog Devices

Companies on stage

Vijay Gadepally, Senior Staff at MIT Lincoln Lab; Co-Founder at Radium

Radium runs a cloud service that serves and fine-tunes open AI models for companies through an API, on its own NVIDIA-chip capacity.

ProjectNandaRamesh Raskar, Founder ProjectNanda; Assoc. Prof.
Massimiliano Versace, Global Head of Emergent AI

Analog Devices sells the radiation-screened ADRF6780S-CSL microwave chip for satellite and space-payload developers.

Signals from this session

4:15 in the recordingVijay GadepallyRadium's data-center software enabled clients to double bandwidth without expanding buildings. Radium's software dynamically adjusts data-center operations, and client deployments doubled bandwidth within the same physical buildings. The efficiency gains allowed companies to defer hundreds of millions of dollars in capital expenditures.
5:14 in the recordingTycho.AI is developing flexible autonomy for high-speed, low-altitude drones. Tycho.AI is building drone intelligence capable of enabling flight at approximately 200 miles per hour and 20 feet above ground level.
5:14 in the recordingTycho.AI is developing flexible autonomy for drones flying at up to 200 mph and 20 feet above ground. Tycho.AI is building drone-intelligence systems designed to enable high-speed, low-altitude flight at approximately 200 miles per hour and 20 feet above the ground.

Members see another 17 signals from this session. To read them, subscribe or sign in.

Viewpoints from this session

3:53 in the recordingVijay GadepallyData centers are operating very inefficiently when viewed simply as buildings.
4:00 in the recordingVijay GadepallySoftware can dynamically optimize data-center operations and unlock substantial efficiency gains, including more bandwidth from the same physical building.
4:23 in the recordingVijay GadepallyImproving data-center efficiency is essential to accelerating AI development, and efficiency should be considered alongside development speed.

Members see another 10 viewpoints from this session. To read them, subscribe or sign in.

Other sessions

12:00PanelRemote ControlChris Rodley, Seaweave · Sidney McLaurin, Gobo Robotics · Rags Gupta, ASYLON · Eric Ragan, Goddard · moderated by Paul Baier, GAI Insights12:00SessionStartup ShowcaseEunice Juarez Garcia, Vitalion · Manoj Bharadhwaj, Atri AI · Peter McHale, Gaia AI · Koneshka Dey, Neobotics Foundation · Malay Kundu, Vizi AI · Hür Köser, Scalables · Matteo Rinaldi, Zepsor Technologies · Hiten Sonpal, RISE Robotics · Ansh Mehta, Neobotics12:00TalkScaling Robot EvaluationRaul Romero, Kite ML12:00WorkshopIn the Loop with DeepMindPeter Danenberg, Google DeepMind12:00SessionMomentumSertac Karaman, MIT · Ramesh Raskar, MIT Media Lab · hosted by Dave Blundin, Link Ventures12:00SessionStudioRaul Romero, Kite ML

Full program →

Members read every signal and every company file.

They also get the Frontier Tech Digest every morning and can ask their own questions. What Frontier Tech Intelligence is →

Subscribe · $99 a month
Independent coverage, not affiliated with any organizer.
↑