Rajat Aggarwal

Co-Founder & CEO at DreamVu AI

Physical AIComputer Vision3D Scene UnderstandingRoboticsComputational OpticsDeep Tech Entrepreneurship

About

I'm Rajat Aggarwal, Co-Founder and CEO of DreamVu AI. My journey began in deep research at IIIT Hyderabad, where I focused on computational cameras and 3D reconstruction, eventually leading me to build a company that bridges the gap between the digital and physical worlds. At DreamVu, we are pioneering omnidirectional 3D vision and Physical AI data infrastructure to solve the 'data bottleneck' that currently holds back humanoid robotics. I am deeply passionate about spatial intelligence and moving beyond the limitations of standard 2D simulations. I'm currently looking to connect with PhD-level researchers and simulation engineers who want to help us close the Sim-to-Real gap. Whether you're working on humanoid training pipelines or retail automation, I’m always open to discussing how high-fidelity 3D data can transform the future of Embodied AI.

Networking

What I can offer

›Expertise in 3D vision and spatial intelligence
›Insights into global deep-tech operations (US-India-Korea)
›Access to high-fidelity multi-view datasets (PRISM/DAIR)
›Mentorship in medical application prototyping and computer vision

Looking for

›PhD-level Research Scientists in VLA/3D Vision
›Simulation Engineers experienced with NVIDIA Isaac Sim
›Partners working on humanoid training pipelines and world models

Best fit for

Robotics researchersAI simulation engineersDeep-tech investorsRetail automation specialists

Current Interests

Physical AI & RoboticsSpatial IntelligenceRetail AutomationVision-Language-Action (VLA) modelsSim-to-Real transfer

Background

Career

Transitioned from a research-heavy background at IIIT Hyderabad and LV Prasad Eye Institute into deep-tech entrepreneurship, founding DreamVu AI in 2018 to lead global operations and innovation in 3D vision.

Education

MS in Computer Science (2016–2017) and BTech Hons in Computer Science (2012–2016) from IIIT Hyderabad.

Achievements

›Holds 30+ patents related to optical mirror designs and omni-stereo light capture.
›Achieved a 66.6% reduction in error rates for VLM models fine-tuned on proprietary data.
›Developed the ALIA 360° camera.
›Recipient of Google India Research Grants (2016).

Opinions

The primary bottleneck for robotics is not model architecture, but the lack of diverse, high-fidelity, real-world 3D data.
Standard 2D images and synthetic simulations are insufficient; robots fail because simulations approximate physics too loosely.
Advocates for open-source datasets to help the community close the Sim-to-Real gap.