Anirudha Ramesh

The Birth of a New Era of Distributed Collective Intelligence

Flexible compute, redefined.

Background Everything: The Missing Predictive Coding and Caching Layer in AI

We Solved Execution. Now What?

AI solved execution. The bottleneck now is purpose.

Should Robots Be Able to Synthesize Aids on the Fly? Tool Synthesis as a Regularizer

Exploring how on-demand tool fabrication can regularize robot learning by transforming rare, difficult tasks into familiar interactions—reducing the combinatorial complexity of open-world robotics.

Let there be cute robots! The sillier path for the creation of data flywheels in robotics

What if instead of billion-dollar autonomous driving approaches, we could kickstart robotics data collection with cheap, cute collectible robots that consumers actually want to interact with?

Learning to Detect by Learning to Predict

The ability to identify and localize objects in an environment by humans and other animals is powered by the agent’s ability to predict where an object is likely to be in the environment. We explore the use of this idea as a prior to improve detection of objects in the scene, which can be seen as the conditional posterior.