As part of our focus on AI, we have talked with Sushantam Mohan, Deputy General Manager & Data Science and Analytics Lead at Mahindra Group.
What is your current role and responsibilities?
I am currently part of the Group Technology Office at Mahindra Group which is the internal consulting team at the group level. I lead data science projects across group companies of Mahindra. The projects can range from marketing, risk, and operations analytics across manufacturing, auto, BFSI, and hospitality.
What inspired you to get involved in the tech industry?
The Tech industry gives you the kick of creating something out of thin air. The ability to drive innovation and work on your digital prototypes is almost unlimited and stretches your imagination.
This industry is also very fast-paced and has some of the smartest minds working which gives you an additional kick to work on the most relevant and cutting-edge problems impacting society.
Can you tell me about your journey?
The journey started purely by chance. I graduated with a degree in engineering and was exploring career options during my college placement activities. I was hired by an analytics start-up and was introduced to the world of numbers and using data to make decisions. I then moved on to HSBC Bank where I spent more than 3 years before moving to Citi where I learned and honed my analytics chops.
I wanted to work for the Indian market given the diversity and sheer dynamic nature of this market. With that in mind, I moved to Mahindra where I got to work beyond my core industry of BFSI across multiple sectors and some of the most interesting use cases.
Why did you choose to work in Data Science and AI/ML?
The entire idea of putting your mind into data and solving the problem at hand proved to be very challenging with immense job satisfaction. Data science gives you the power to look at the complete picture and helps unravel the mysteries and solutions which are not visible to the naked eye. I was always fascinated with numbers but never imagined training a computer to look at trends and fitting models will become such powerful tools.
What is your favorite part of your job?
My favorite part of the job is the variety it offers. One day I work on a credit risk model, moving onto marketing, and then try my hands at operational analytics. I have a view across the entire lifecycle of the product.
Market surveys to design, customer segmentation to target, A/B testing to sell, engagement models to service and retain. No other job probably other than the CEO’s job offers such wide scope to make a difference.
What are the dangers of AI and how can they be prevented?
I find a couple of dangers that are quite discomforting to me. Data science the way it’s used can accentuate existing biases in society. Since most of the models are backward-looking, they fit basis what data “shows” them and if the data has some inherent biases it will seep into the model and it would seem even machines are biased.
AI can also create a polarized society as personalization models can show only things someone is interested in and completely shut out information that does not conform to the beliefs of the user.
Is it possible to eliminate or at least reduce the risks of bias in AI-driven technology?
Eliminating the risk of bias is not possible. We can reduce it by being cognizant of the same and taking remedial actions. There are plenty of ways to do the same. Picking right data, rightsizing of the data, dropping, or truncating values or attributes that have known biases, masking the data to make sure machines doesn’t “see” the biases.
In the end, is the person who is building the model needs to be very careful in picking fields and suppressing values that are known to create biases. Unknown biases are difficult to identify and eliminate.
Similarly, can we prevent AI from invading people’s privacy?
AI depends on data and restricting the data to maintain privacy would mean no AI. Hence there must be a fine balance between the use and boundaries that need to be drawn. Usually, most of the AI models operate at a macro level and don’t snoop down to the micro-level although people may feel it is.
Anonymization and tokenization are gaining immense popularity for privacy reasons. Using these we are masking the customer information but continue to capture the trends which help both the user and the companies for better and efficient targeting.
What would you suggest to develop safe and trustworthy AI?
One of the bedrocks to have safe and trustworthy AI is to make sure the system has the necessary checks and balances to self-correct and improve. Modeling teams should aspire to have a maker–checker model with both teams having AI experts. If one team is building models others should check for inherent issues, biases, and misuse of information.
Strong compliance policies and a system open to debate and feedback are key for AI to operate in a sustainable way. Models should be periodically reviewed, retired, or retrained to make sure they remain compliant.
Do you believe that artificial intelligence is the future of technology?
No, AI is part of technology and not the technology itself. AI is there are aid human beings in taking decisions in a more objective way, automating mundane tasks, drive faster decision making, and enabling a more comfortable life for everyone. It must co-exist in the overall scheme of technological advances in manufacturing, services, and operations.
Technology will always be an overarching phenomenon with AI as one of the most important spoke in the wheel of change.
Do you have a memorable story or an anecdote from your experience with AI you’d like to tell?
During one of my visits to a remote rural town in India, I was interacting with a layman who had never heard about AI. I was trying to explain to him how AI can help him in making decisions about loans applications much faster and would rarely give him wrong recommendations.
The fellow was not very impressed. He said that machines can never replace the human experience. To convince him I agreed with him. I told him that AI is nothing but a collection of human experiences, just that everyone learns from their own experiences, but AI is sum total of the learning experience of all humans who have ever processed a loan application. That made that fellow trust me and try our algorithms.
Finally, do you have any advice for aspiring testers who want to grow in the Data Science field?
Testing is a tough job and an immensely critical one. It’s like electricity, people may not see it working but without it, the product will be useless or incomplete.
The only advice I have for testers is to stay agile, look for issues constructively and learn to engage with developers. The idea is to make the product better and ultimately serve the end-users in the best way possible.
Focussing on the big picture always helps without losing sight of the work at hand.