Sreetama Das has worked in various businesses, including manufacturing and healthcare, as a researcher in Computational Biology and Structural Biology/Biophysics and as a Senior Data Scientist. INDIAai caught up with Das, who works at GSK as a Senior AI ML Engineer, to get more about her thoughts on AI.

Graduating with a physics degree and pursuing research in computational biology is a rare combination. Could you tell us about your professional journey?

Firstly, I would like to thank Dr Nivash Jeevanandam and INDIAai for this opportunity to share my professional journey and inspire young minds. As a high-school student, I had always dreamt of becoming a scientist and developing new solutions to help improve people's lives in whatever way possible. Then I came into touch with X-ray crystallography during my postgraduate studies in physics and fell in love with it. There are exciting stories about the applications of X-ray crystallography - the discovery of the structure of DNA and insulin, to name a few. It is an application-oriented and interdisciplinary domain, and I chose to pursue my PhD therein. This domain was a transition from physics to biophysics and computation. My PhD research revolved around crystallographic techniques on protein molecules to derive their structure and obtain insights into the function of those molecules by analysing the sequence, structure and dynamics information using various biophysical and computational techniques. I ended up relearning some aspects of biology, as well as learnt about biological molecules from textbooks as well as research papers. I also picked up coding skills and machine learning techniques during this time. This experience, in short, was my journey into computational biology. 

Could you tell us about your typical day as a Senior AIML engineer?

Sure. As I work in an MNC, each project usually collaborates with many teams having varying expertise. The early phases of a project would involve a lot of discussion among the involved team members to frame the problem, understand the availability and type of data sources, and test the feasibility with simpler machine learning algorithms. Once feasibility is proven, the team, led by the senior AIML engineer, will look at methods to further improve the solution and deploy it. So I typically spend considerable time discussing methodologies and next steps with the team members and programming. I also try to keep aside some time for learning and development.

Choosing a research area and defining the research problems are critical stages in the research process for any researcher. Could you tell us about the research phase?

Researchers themselves usually choose the high-level research area. However, it is a good idea to identify broad themes of interest to the would-be researcher, either based on the curriculum or from reading general science articles (from magazines and newspapers), to form some idea about the research across the world. Doing short-term research projects is another way to learn more about research topics. It is also important to network - have conversations with senior researchers or professors to get some idea about what topics could be interesting enough to work in for the next 5 - 10 years. 

The next step is to identify appropriate research labs where one intends to conduct the research based on information available on lab pages and from informed seniors/professors/ teachers. The research theme of the lab should match the researcher's interests. For example, some labs require a research proposal at the outset, whereas others decide on the exact research problem after reading, discussion, and assessment iterations. Developing an ability to read research papers to identify the gist is helpful in this context.

How was your transition to AI and machine learning as a cross-disciplinary researcher? How did you adapt to programming and artificial intelligence/machine learning concepts?

My research phase was very interdisciplinary and enriching. I learned how to frame the problem statement better through hands-on experience and a literature survey. I also became a lifelong learner, discovering new methods based on what was needed to solve the problem. Although my post-graduation was in physics, I needed to relearn concepts of chemistry and biology to delve deeper into the problem statement and solve it. In addition, I picked up programming skills and learned to implement machine learning techniques through the problems I solved. By the end of my PhD, I realised I had gathered several of the skills of a data scientist and chose to transition to an industry position in data science and machine learning.

You are an inspiration to many professionals, particularly women, who wish to pursue careers in AI/ML. Who inspired you during your research and career transformation?

My teacher, Prof. Aloke Kumar Mukherjee, inspired me to pursue a research career. I also heard of great female researchers like Rosalind Franklin and Dorothy Hodgkin, who had worked in Crystallography, their stories of grit and perseverance. These helped me persevere, especially when research projects did not work out as expected (which could be true a lot of the time, depending on the field of research!).

I planned for a transition towards the end of my PhD. Unfortunately, back then, I did not have any role models. Still, a reputed science journal had published an article about data science being one of the most coveted jobs in this century and how researchers would have developed some of the relevant skills through their journey. So I took the plunge and started preparing for and giving interviews. I saw other researcher colleagues with somewhat different backgrounds successfully transition during this time. That gave me the courage to persist with my efforts.

India's adoption of AI across various sectors has been accelerating. What is your assessment of the growth of AI in India over the next five years?

Yes, we have seen several startups, from healthcare to fintech, edu-tech, retail and e-commerce, which use AI in one way or another for particular problem areas. At the same time, larger companies continue to look for AI applications in their traditional domains and venture into new fields. As a result, we can see an increase in opportunities in this field over several years, leading to sustained demand for trained professionals.

What, in your opinion, is a necessary skill for any professional or student interested in pursuing a career in AI/ML?

The field is very dynamic. What was once trending can become obsolete in a few years. So a necessary skill would be learning to adapt - learn, unlearn and relearn as required.

According to researchers, AI has accelerated the development of emerging technologies such as big data, robotics and IoT. But is it necessary to introduce AI to schoolchildren through their subjects?

It is essential to focus on the fundamentals at the school level - logical thinking and problem-solving skills. School children have already get introduced to computer programming. Some simple techniques like linear regression and clustering as part of curricula - may be called out explicitly as machine learning techniques in computer science courses. General reading about science and machine learning can be encouraged to know more about exciting applications. Statistics is also essential. These can form the foundation for more detailed courses in AI at the college level. 

Could you recommend any research papers or books that every AI/ML student should read?

I would recommend a few to help get started, which should be supplemented by application-area specific reading. Elements of statistical learning (Hastie and Tibshirani) and machine learning - a probabilistic perspective (Kevin Murphy) are good to start with education. Then, for deep understanding, follow the work by Ian Goodfellow and Geoffrey Hinton. Other options would be blogs by Google, Facebook, and Microsoft, exciting reads on medium.com, analytics India magazine, analytics vidhya, etc.