ODSC Video from Meetup in Bengaluru, India

Meetup in Bengaluru

Presented By Anand Subramanian and Dipanjan Sarkar
Anand Subramanian and Dipanjan Sarkar
Anand Subramanian and Dipanjan Sarkar

Anand Subramanian is rated among the top 10 data scientists in India. A Gold medalist from IIM Bangalore. Alum of IIT Madras, BCG, Infosys Consulting & IBM. Geek to the core.

His affair with programmatically analyzing data began in the mid-90s at IIT Madras. From Linux based grep, sed and awk, he then moved on to PERL and Excel, and now Python. "Tools are never the concern, the motive and the execution is," believes Anand.

Dipanjan (DJ) Sarkar is a Data Scientist at Red Hat, a published author, and a consultant and trainer. He has consulted and worked with several startups as well as Fortune 500 companies like Intel. He primarily works on leveraging machine learning and deep learning to build large- scale intelligent systems. Having a passion for data science and education he also mentors people and organizations like Springboard and acts as the editor and key contributor for Towards Data Science an online publication dedicated to Data Science and AI.

Dipanjan's interests include learning about new technology, financial markets, disruptive start-ups, data science, artificial intelligence, and deep learning. In his spare time, he loves reading, gaming, watching popular sitcoms and writing interesting articles.

Presentation Description

Talk 1: Interpreting Machine Learning

Machine learning algorithms are increasingly black-box models. However, their outputs are business data that humans need to understand and act upon. For example, if a clustering model suggests 4 customer clusters, how do we identify and characterize these? If a random forest model suggests a pattern of classification, how do we understand the dominant factors and the irrelevant ones?

These topics fall under the umbrella of model visualization -- where the inputs, process, and output of machine learning models are the topics of understanding. This talk explores some of the prevalent ways of visualizing machine learning models.

Talk 2: A brief peek into interesting real-world applications for AI

Thanks to better compute and storage, we have been finally able to dive into the realm of building applications powered by machine learning and deep learning. Solving real-world problems in the industry is slightly different from theoretical concepts or papers backed by research on standard datasets. Problems start cropping up including noise in data, lack of good quality data, sources of bias, class imbalance and domain-specific issues.

This talk will cover some interesting case-studies of problems I have solved in the past or which I'm trying to solve in the present, leveraging some crazy ideas from machine learning, deep learning, and my personal intuition. Case studies we will cover include:

- Generating insights on enterprise incident data with NLP

- Predicting Device Failure with Deep Learning

- Detecting potential anomalies in Infrastructure

- AI-Based Application Insights for Developers

Presentation Curriculum