Sujoy Roychowdhury is a Group Manager and Senior Data Scientist in IBM and leads the Recommendation Systems and Deep Learning practice in IBM’s AI services unit. He has 16 years of experience in Analytics and Data Science. He is a recipient of Major Outstanding Technical Achievement Award - the highest individual technical award in IBM and has been a Best of IBM awardee in 2019 – given to IBM’s top-performing 1000 employees globally. He has filed 3 patents, co-authored two HBS cases and has 7 international conference publications, including two best paper awards. He has been a visiting faculty at IIMs and given talks in various industrial and academic conferences. Sujoy is an Electrical Engineer and did his EPGP from IIM Bangalore in 2012-13 and is currently doing his Ph.D. on speech recognition AI systems.
Building of AI systems that work in production has many additional challenges from that of building one-off machine learning models. Speaker will focus on two major aspects of building AI systems.
Model maintenance involves not only the IT aspects of maintaining a production system but also having a system to continuously monitor performance and evolve as the trends in data change. We will discuss some of the challenges in this problem and look at some methods commonly used for this problem.
The model deployment consists of deploying large scale industrial models and the options available in them. Speaker will talk about the choices available for model deployment and how those choices are influenced by use cases. We will also talk about deploying models to edge devices and what are the challenges regarding the same.