Shubhabrata Roy

Shubhabrata Roy

Senior Data and ML Consultant at Sioux Technologies

What is impossible today becomes feasible tomorrow thanks to data. Machine learning, particularly deep learning, has achieved advances in recent years, transforming fantasy into reality. These technological improvements are extraordinary, yet they rely on massive volumes of data being accessible. Furthermore, having data is insufficient for the type of machine learning called supervised learning. While powerful, supervised machine learning requires data with labels that are used to train and direct the algorithm as it learns.

Unfortunately, data does not come neatly organized with labels in the actual world. Enterprises acquire vast volumes of data, yet only a small percentage (if any) of it gets annotated. Attempts are made to manually annotate them to harness supervised learning potential, but this activity may be expensive, inefficient, and time-consuming. In this I will talk about how to train your machine learning models with limited labels. 

Learning with limited labels

Shubhabrata has been in the AI domain for almost 10 years and led multiple cross functional teams to develop many successful products. Shubhabrata holds a PhD in Signal Processing from École normale supérieure de Lyon, a French grande école. After PhD Shubhabrata joined an award-winning start-up in Budapest as a senior Data Scientist for two years and worked on emotional intelligence. Since 2016 Shubhabrata has been living in Eindhoven and worked for companies like HERE Maps, Huawei Research in different capacities on AI applications for smart mobility. Currently Shubhabrata is working at Sioux Technologies as Senior ML Consultant on health care analytics and large language models. Shubhabrata also loves talking about technologies and gave a talk on AI for mental health at TEDx RWTH Aachen in 2022. 

Agenda Talks


Data, AI & ML Con

11:40 to 12:25
02 Nov 2023

Learning with limited labels

What is impossible today becomes feasible tomorrow thanks to data.