New Delhi-based startup Neocambrian AI has launched India’s first robotics data factory, focused on building large-scale human action datasets for Physical AI and embodied intelligence systems. Founded by Abhinav Kukreja (who previously founded DataVantage), the company aims to fill a critical gap: robotics currently lacks the massive, high-quality datasets that have powered advances in large language models.
The data factory uses advanced capture technologies—including egocentric video systems, motion tracking hardware, and stereo capture rigs—to create a high-fidelity, pre-training scale database of human actions. This data is essential for training vision-language-action (VLA) models and world models that enable robots to understand and interact with real-world environments.
Neocambrian AI positions India as a potential global leader in Physical AI datasets due to its vast workforce, diverse real-world environments, and experience in distributed services. The startup recognizes that as AI moves beyond chatbots into robotics and physical systems, data infrastructure for real-world learning becomes the critical layer of the next AI wave.
The launch comes amid growing momentum around Physical AI and increasing scrutiny around privacy, worker consent, and ethical data collection practices in human activity datasets.
FAQs [Frequently Asked Questions]
Q1: What is Neocambrian AI’s robotics data factory?
A1: It’s India’s first facility creating large-scale human action datasets using egocentric video and motion tracking for training Physical AI and robotics systems.
Q2: Who founded Neocambrian AI?
A2: Abhinav Kukreja, an entrepreneur who previously founded DataVantage, established the startup to address robotics’ lack of training datasets.
Q3: Will researchers access this data?
A3: Yes, Neocambrian AI will provide thousands of hours of collected human action data free to Indian researchers working on VLA and world models.