Developing visual foundation models that synthesize 3D geometry and physical dynamics into actionable intelligence for complex, unstructured environments.
High-fidelity spatial reasoning & multi-modal integration.
A city-scale intelligence initiative for high-density environments. Dense World models crowd flow, mixed mobility, and infrastructure constraints to generate robust forecasting and planning signals where standard low-density assumptions fail.
A Joint-Embedding Predictive Architecture optimized for the extreme density of South Asian urban environments. FactorJEPA learns invariant representations of world dynamics by factoring geometry, semantics, and temporal flow into discrete, manageable latent spaces.
Our Vision-Language-Action foundation. Featuring a novel Locomotion-Aware Chain-of-Thought (LA-CoT) mechanism, PragyaVLA bridges high-level linguistic reasoning with low-level torque control for dexterous robots in domestic and industrial settings.
A framework for fall-resilient humanoid motion inspired by Kalaripayattu, the world's oldest martial art. KalariSena uses high-frequency motion tracking to enable autonomous systems to recover from physical disruptions and navigate precarious terrains with biological fluidity.
The regulatory and safety backbone of the ecosystem. Emphasizing deterministic alignment, safety guardrails, and AI policy, the Kalam Protocol ensures that autonomous systems operate within ethical boundaries across mesh-networked infrastructures.
Cross-ecosystem preview of raw sensory input versus processed spatial understanding.
Raw Input
FactorJEPA Map
FactorJEPA successfully reconstructing 3D volumes in 98% occlusion scenarios in dense traffic.
Kinematic Input
KalariSena Policy
PragyaVLA + KalariSena achieving 100% stability on uneven simulated and real hardware surfaces.
We are selectively opening our spatial intelligence frameworks to global research institutions and industry leaders.
BITS, Goa Campus, India
