Swetha Rajkumar

About

Swetha Rajkumar is a rising researcher at the forefront of robotic manipulation and embodied AI. Her primary research focuses on enabling robots to operate in open-world environments, tackling the critical challenge of generalizing to novel objects without extensive robot-specific data. Rajkumar’s most notable contribution is the development of KALIE (Keypoint Affordance Learning from Imagined Environments), a groundbreaking framework that fine-tunes pre-trained vision-language models for robotic manipulation. KALIE uniquely leverages imagined environments and keypoint affordances, allowing robots to learn manipulation skills from simulated or imagined scenarios rather than costly real-world robot data. This approach significantly reduces the data burden and enhances a robot’s ability to handle unseen objects, marking a major step toward truly generalist robotic systems. Her work has already garnered attention, with her KALIE papers accumulating citations in the short time since their release. Rajkumar’s research sits at the intersection of computer vision, natural language processing, and robotics, and her innovative use of foundation models for embodied tasks positions her as a key voice in the next generation of AI-driven robotics.

Research Focus

Key Achievements

Top Papers

1. 1
   KALIE: Fine-Tuning Vision-Language Models for Open-World Manipulation Without Robot Data

   2 citations · 2025
2. 2
   KALIE: Fine-Tuning Vision-Language Models for Open-World Manipulation without Robot Data

   2 citations · 2024

Key Collaborators

Contact & Links