Today is Episode 4 of Interview series on Expert-talks @MAAVRUS,  with Leaders in the Analytics, AI and Transformation space.  For this episode, our CEO Mahadevann Iyerr (Mahaa) is in conversation with Rahul Pednekar, Vice President & Head – Advanced Data Analytics, Actuarial and Data Insights at Swiss Re. Prior to Swiss Re, Rahul has had a wonderful career at Vodafone, JP Morgan Chase and Infosys.  He also frequently participates in AI & Analytics hackathons, and speaks at Industry conferences on the Applications & Future of AI.

We are sure listeners will greatly benefit from Rahul’s depth of knowledge, and his thoughts on achieving analytics success, by using a practical project lifecycle.  We are listing below, a few key points from the interview :

• According to Rahul, Data Science is an intersection of Programming, Maths & Statistics and Domain understanding, and to be successful one needs to have a well-rounded ability across all these aspects.
• Data scientists cannot learn in isolation. They should work and collaborate with other data scientists to learn more.  Open hackathon platforms are a great place to ask, share and learn. Also when participating successfully in hackathons, it is always a good idea to create a digital footprint of your approach through blogs etc. Apart from getting a sense of validation, it could also inspire fellow data science professionals to try & learn.
• To build domain knowledge, one should stay updated by reading industry articles and magazines and by interacting with business domain practitioners and learning from their viewpoints.
• From a data scientist’s perspective, it is also necessary to spend time learning and building models on their own, so as to appreciate any industry-specific nuances,  for eg for actuarial modelling R is better suited than Python.
• When embarking on an ML transformation project, it is always good to build an initial prototype to check for business alignment on the controllable input factors, and possible outcomes, as well as get a sense of the constraints from a data availability & quality perspective.
• External data is very important and helps build a market perspective for the business. However in many cases, real-world data may not be available, so one has to look at trusted sources for forecasted data about possible input factors. For eg when it comes to global economic or financial data one could explore sources like Moody’s, Bloomberg etc.
• Data scientists will need to be able to build models, where they can help business understand in a simple manner, why a model is predicting in a particular way, and also justify how the outcomes predicted by the model is in alignment with stakeholder expectations/business case.
• For analytics / ML projects to be successful, Data scientists need to have a larger understanding of data engineering, data ops, model building and endpoint creation through ML ops. Otherwise, the model is likely to stay just on their Jupyter Notebook, and not get operationalised.
• Organisations are still trying to understand the possibilities and limitations of OpenAI and deep learning platforms. While there are a lot of innovation and business transformation possibilities with Open AI, there are still concerns from an ethical viewpoint, unintended bias, computing capability and investments.
• Directionally the future of AI will be a collaboration between the ingenuity of humans and the computing scale of machines.
• Young aspiring data scientists will need to stay curious and hungry for continuous learning, build a strong foundation in statistics, and collaborate extensively with data scientist communities, both within and outside the company.