My recent past.
I recently graduated from The Australian National University(ANU) with a Masters degree in Computer Science, specializing in Artificial Intelligence. My current research interests are in scalable directed exploration for Reinforcement Learning, and AI safety.
Back in 2008, I joined Infosys Ltd. as a software engineer after completing my Bachelors in Electronics and Communication engineering. The next six years was spent designing and developing software for a wide range of domains. Towards the later years at Infosys, I worked on a proof of concept project that used a Machine Learning algorithm (Fisherfaces) to perform gender classification. This project introduced me to the world of Machine Learning, and subsequently Artificial Intelligence (AI). Smitten by the promise of AI, I resigned from my job to pursue a Masters degree in Computer Science.
During my Masters, I participated in the Logic Summer School, a week long exposition into the recent advancements in Logic. As part of the program I led a team to build an online platform to enable Logic students to write, verify, and store System L style Natural Deduction proofs for propositional logic. The project, dubbed Proof Assistant, was completed under the supervision of Prof. John Slaney.
Later I joined Prof. Marcus Hutter’s Intelligent Agents research group to work on my Masters research project. In this project we introduced a novel optimistic count-based exploration algorithm to perform directed exploration in large Reinforcement Learning problems. Our work was later published as the paper ‘Count-Based Exploration in Feature Space for Reinforcement Learning’ at the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI). My Master’s thesis provides an exposition of this work.
Below you can see our agent in action, playing the Atari games Montezuma’s Revenge and Venture.
A full list of my professional projects can be found on my AngelList profile, and my personal projects can be found on Github. The best way to contact me is by email: sura[email protected]. Feel free to follow me on Twitter, connect with me on LinkedIn, or browse through my CV.