Krishna’s research aims to deeply intertwine "classical" approaches to computer vision and robotics with modern machine learning. While deep neural networks have led to tremendous advancements in perceiving and interpreting 3D scenes, robotic systems of the future need flexible, interpretable, and actionable representations of the world they inhabit. Krishna’s work leverages insights from multiple communities (vision/robotics/graphics) to drive meaningful progress towards these grand goals. Use cases of his work include designing intelligent robot assistants, immersive virtual environments, safer autonomous vehicles, and video understanding systems.