Bhavika Devnani

Papers

Total Citations

H-Index

About

Bhavika Devnani is a leading researcher in embodied AI, with a focus on advancing zero-shot generalization for robotic navigation. Her most influential work, "ZSON: Zero-Shot Object-Goal Navigation using Multimodal Goal Embeddings" (2022, 41 citations), introduces a scalable method for open-world ObjectNav—enabling virtual agents to locate arbitrary objects (e.g., "find a sink") in unseen environments without task-specific training. This approach eliminates the need for ObjectNav reward engineering, leveraging multimodal embeddings to bridge visual and semantic understanding. Devnani’s contributions are pivotal for developing robots that can operate in unstructured, real-world spaces, pushing the boundaries of sample efficiency and generalization in embodied AI. Her work has garnered attention for its practical implications in home robotics and autonomous exploration, earning citations from top venues like CVPR and NeurIPS. By tackling the core challenge of zero-shot transfer, Devnani is shaping a future where AI agents can intuitively navigate and interact with their surroundings, making her a rising voice in the intersection of computer vision, natural language processing, and robotics.