Max Robotics — Global Service Robot Directory + US Certification

About

Anurag Ajay is a leading researcher at the intersection of robotics, machine learning, and embodied AI, whose work bridges the gap between simulation and real-world robot learning. His key contributions span three interconnected areas: developing hybrid physics simulators that combine analytical models with stochastic neural networks for more accurate contact dynamics, advancing state estimation through discriminative learning approaches like Backprop KF, and pioneering embodied question answering with OpenEQA. His most influential work, "Backprop KF" (102 citations), introduced a novel framework for learning discriminative deterministic state estimators that outperform traditional generative models, particularly for rich sensory inputs like camera images. His work on augmenting physical simulators with stochastic neural networks (95 citations) has become foundational for robot planning and control under uncertainty, enabling more reliable sim-to-real transfer. Ajay's research on reset-free guided policy search and endoscopic capsule robot navigation demonstrates his commitment to practical, real-world applications. His recent OpenEQA benchmark (2024) is shaping the next generation of foundation models for embodied AI. With over 300 total citations and publications spanning top venues like NeurIPS, RSS, and ICRA, Ajay continues to push the boundaries of how robots learn, perceive, and interact with complex environments.