Building intelligence
for the physical world.
Sagvad is a Physical AI research lab. We work on robotics, world models, and the core problems of deploying learned systems in the real world.
Currently
What the lab is thinking about
The gap between AI that reasons about the world and AI that competently acts in it remains the field's most consequential open problem.
We are spending this season on the perception–action loop in unstructured environments — what representations carry through it, where simulation misleads, and how to evaluate honestly when the test set is the real world.
Writing
Recent notes from the lab
Research Notes
India AI Impact Summit 2026: A Defining Moment for Global AI
From the Delhi Declaration to $250 billion in commitments, India's landmark AI summit brought together world leaders, tech giants, and researchers to chart the future of inclusive, impactful AI.
AI Safety
Safety Beyond Slogans: Practical Frameworks for AI in the Real World
Bias, fairness, transparency, and accountability for AI systems — including the embodied systems that act in the physical world, where mistakes have force, velocity, and consequence.
Perception for Action
Perception for Action: Lessons from Computer Vision in Manufacturing
A look at vision systems deployed in manufacturing — quality control, defect detection, robotic guidance — and what those deployments teach us about perception that has to drive action.
Working in this space, or thinking about it? We'd like to hear from you.
Get in touch