Use case in production
Reinforcement learning for real-life continuous spaces
Principal Data Scientist, Michelin
Ameya Divekar is passionate about leveraging the potential of deep reinforcement learning for building autonomous systems in his role as a Principal Data Scientist with Michelin. Particularly, he is focusing on building intuitive debugging tools for deep reinforcement learning so that data scientists get clarity of the several varying parameters of a deep reinforcement learning construct. Previously, he has worked in technical leadership roles across several companies such as John Deere, Dassault Systemes and PTC. He is creative at heart and takes pride about his passion for innovation, backed by 4 patents and 3 trade secrets in his career. He holds a Masters Degree in Mechanical Engineering from Clemson University and more recently Diploma in ML & AI from IIIT Bengaluru.
For environments with optimization of real-world multi-dimensional tabular datasets, such as optimization of chemical process parameters, it is complex to do the same that is done for virtual game environments i.e. rendering game screens. The challenge is exacerbated when state and action spaces are continuous in nature and dimensionality is high.