Inspired by Mata Amritanandamayi Devi, the world-renowned humanitarian known as Amma, Vishnu began to dwell on finding solutions to problems in educational equity while still in 2nd grade. An interest in computational thinking and design thinking led him to discover the $15 million Global Learning XPRIZE competition, which challenged participants across the globe to develop software that teaches reading, writing, and arithmetic for kids with limited or no access to teachers or the Internet. Vishnu was the first to join a new Carnegie Mellon University team, RoboTutor, in May 2015 to compete in the XPRIZE competition. As chief technology officer (CTO), Vishnu developed handwriting recognition, text-to-speech (TTS), computer vision, and machine learning capabilities designed to maximize learning gains in kids. RoboTutor advanced to the semifinals in June 2017, and in September 2017 was awarded $1 million as one of five Global Learning XPRIZE finalists among the original 135 contesting teams worldwide. Since December 2017, RoboTutor has been deployed in Tanzania for hundreds of kids to use, with the potential of raising the literacy level of 250 million children worldwide and lifting them out of poverty.
| E | vrt@ieee.org |
| T | @vrtejus |
| W | +1 (650) 516-6550 |
| O | by appointment |
| L | https://www.linkedin.com/in/vishnutejus/ |
The RoboTutor project has highlighted the need for bringing together artificial intelligence (AI) and machine learning, immersive technologies such as augmented reality (AR), and cognitive technologies such as emotion detection in one platform. To create intelligent systems that can better understand the real world, I have been working on AI that can reason and learn through exploration, a fundamental trait of a system with general intelligence, to get answers. Deep learning approaches have been successful in solving recognition tasks such as image recognition, sentiment classification, and basic video prediction. However, these systems require large amounts of labeled training data, which is expensive, time-consuming, and is tricky for domains without human expertise. Furthermore, human-generated datasets have been shown to cap model performance thereby not achieving on-par or super-human performance. Towards solving these limitations of current machine learning approaches, the general direction of my research has been creating systems that can capture the compositional nature of the real world.
Here is my publications list at Google Scholar.
The key areas of my research are machine reasoning for visual question answering, natural language processing, understanding commonsense, and capturing the compositional nature of the real world. I am also interested in building robust and interpretable systems that can generalize for real-world applications.
I am looking forward to collaborating with fellow researchers in AI and reasoning. Feel free to send me an email!
https://vishnurtejus.org/