Arunchandar Vasan

Short CV:

Arun Vasan obtained a B.Tech degree in Computer Science and Engineering from IIT Madras; and MS and PhD degrees in Computer Science from the University of Maryland, College Park. He is currently a Principal Scientist in the Data and Decision Sciences division of TCS Research. His current research interests are broadly in real-time decision making for infrastructure cyber-physical systems with applications in building energy management; renewables; electric mobility; and airline operations. The approach is to design and implement classical optimisation and/or reinforcement learning-based agents for efficient operations of infrastructure systems. The USPs of the work include computational scale that is practical in the real-world; and the ability to work on small-data/limited actuation problems using domain knowledge.

Arun has several granted patents and papers to his credit, including three that won best-paper awards. He has been a referee for several journals and conferences; and collaborator/mentor in many industry-academia interactions. He has served as a guest faculty at the Department of Computer Science, IIT Madras. He has also contributed to research thinking at colleges in and around Chennai. He has been recognised by TCS as a Distinguished Scientist.

Title of Talk 1: Decision making for physical systems – An industry perspective

Synopsis: With the advent of commoditized sensing and scalable AI/ML-based techniques, the operations of many legacy physical systems and processes (in energy, mobility, etc.) are being digitally reimagined as cyber-physical systems. Newer physical systems such as electric/autonomous vehicles are digital by default. Deriving business value from such systems through efficient operations typically requires intelligent real-time decision making. In this talk, we identify some common pitfalls in real-time decision making for physical systems from an industrial practice perspective. We highlight our attempts to overcome these both using domain knowledge and novel solution techniques such as Physics-inspired machine learning; reinforcement learning; and multi-agent algorithms. We illustrate our approach with case-studies drawn from various industry verticals and conclude with some lessons learnt.

Title of Talk 2: A research career in the industry

Synopsis: Research in academia tends to optimize for quality and quantity of publications. Research in the industry, however, is often a different ballgame where the primary yet typically unstated objective is to have an impact on business beyond publications and patents. Indeed, many problems in the industry may not require original research at all. In this talk, we will first explore the basic elements of the research process common to both academia and industry. Then we will discuss how to fine-tune this process to be successful as a researcher/innovator in the industry. Specifically, we will discuss how one could 1) identify problems from a business value perspective; 2) demonstrate short-term progress even while focusing on longer-term projects; and 3) communicate research to key business stakeholders. We will conclude with some concrete suggestions on operationalizing these ideas

Title of Talk 3: NONE

Synopsis: NONE