Artificial Intelligence (AI) and Data Science continue to lead innovation and technological advancements in today’s world. This is an edited excerpt from an interview of Rohit Pandharkar, Head of Data Science, Mahindra Group, conducted by Amit Paranjape, Chairman of the IT & ITES Committee at MCCIA, as a part of MCCIA’s YouTube series, ‘Decoding Artificial Intelligence with Industry Leaders’. Mr. Pandharkar talks about the applications of AI, the use of data as an asset, and the future of technology. He also explains how the Mahindra Group is using AI to revolutionize operations and how contemporary businesses can make the most of a new era of AI and Big Data.
Almost a decade ago, Tech Mahindra worked on an interesting project for AT and T, one of their largest customers in the US. The brief given to them was to build a system where the gadgets in the house acted like friends. For example, when it’s time for you to come back from work the garage door says, “‘I anticipate you arriving at this time, I will be alert and let you come in whenever your car shows up.” Today it’s called the Internet of Things (IoT) and is one of the hottest trends in the industry, but back then it wasn’t a buzzing topic and was named Social Web of Things (SWOT). Rohit Pandharkar, the then Deputy CTO at Tech Mahindra played an important role in this project.
Mr. Pandharkar is an engineer from the College of Engineering, Pune. In 2009, he did his masters from MIT Media Lab, where he worked under Prof. Raskar in the area of Image Processing. His research focused on a prestigious project funded by the Defense Advanced Research Projects Agency (DARPA), involving laser equipment worth $ 500,000 to build a camera that could detect 3-D images, their material, and size while being able to see around corners and blind spots. In 2011, he met Mr. Anand Mahindra who had been invited as a guest of honour for the silver jubilee celebrations of the MIT Media Lab. This turned out to be a turning point for him as Mr. Mahindra, impressed with his credentials presented him with an offer to join the Mahindra Group.
He grabbed the opportunity and came back to India to work as the Deputy CTO at Tech Mahindra. He also worked with Mahindra Susten where they set up an IoT based platform called ‘Machine Pulse’ which tracked the solar power generation across multiple solar plants. While they had multiple large customers like Infosys and Finolex, the challenge was to compete with the revenue model of global players like Schneider Electric and Siemens who offered this as a free service, with their inverter and other equipment.
Subsequently, he joined HeadSpin, a San Francisco-based company funded by Google Ventures that was an AI-based platform for mobile testing. They worked on techniques to predict when Twitter/LinkedIn is going to crash to give users a seamless experience. In 2016, he came back to the Mahindra group to join as the head of the ‘Centre of Excellence in AI and Data Science’ which they were setting up. This was after the Cambridge Analytica and Facebook scandal were data was used to target people and influence US presidential elections. The world had woken up to the tremendous power of data and the Mahindra management was keen on leveraging data to get key business insights. Initially, it was tough to attract the top data science talent to a traditional economy business like Mahindra but today they have some of the best talent in the form of ex-employees of Citibank, Musigma, and Flipkart and even from universities like Carnegie Mellon and Harvard Business School.
While speaking about the applications of exponential technologies in the Mahindra Group, he explains that they are using AI and ML in three domains. Their first area of focus is to improve the car sales of the Mahindra automotive business by using Artificial Intelligence (AI) and Machine Learning (ML). Simply explained, they use a robot to do micro testing to determine the percentage of marketing budget to be used for platforms like Facebook, Instagram, and Google Paint shop optimization is the second area where the Mahindra group is using AI algorithms. This is a project, driven by Mr. Pandharkar’s colleague and is being implemented at the SUV manufacturing plant at Chakan, Pune. It is one of the first attempts in the Indian automotive industry where sensors will detect humidity, temperature, wind flow inside the paint shop and the algorithms will solve the equation to arrive at the right speed, thickness, and viscosity of paint that should be used for painting the car. This will save a lot of rework and reduce paint wastage, thereby saving hundreds of crores annually for the company. The third area of application is Mahindra Finance, which operates in almost 60% of villages in India is also using AI to give loans to people who do not have a credit history. They have gathered a lot of data over the past 25 years based on their manual credit evaluation process and are now using the same to build machine learning algorithms for credit evaluation. One of the parameters that is being used in the process, is the distance of the person’s residence from the town centre because it affects the ability to collect money. Another parameter being used is the periodicity the person is going for. For example, farmers have a half-yearly cash flow whenever Rabi and Kharif crops harvest and need different timelines for EMI repayment. If a farmer is asked to pay EMI per month, he doesn’t have a cash flow to pay off EMI per month and will show up as an NPA in normal parlance. But for a half-yearly product, farmers end up being their best customers. Taking this into consideration they provide products for monthly periodicity, quarterly periodicity as well as half-yearly periodicity.
In his opinion, these implementations were possible thanks to the leadership at Mahindra which was conducive to these innovations. In most companies, it is a challenge to convince the business head and the internal stakeholders to adopt data science in their area. Hence, it’s important to adopt the ‘Control and Experiment Design Strategy. This means that AB testing of the control and experiment group should be done, where the robot is applied to some places and not applied to some. On keeping all the other attributes the same and comparing the results of both groups, the impact can be attributed to ML.
To conclude, he says that Deepfake identification will be a big area of research in AI in the coming 5–10 years, with fake news causing massive trouble to Facebook and Google. Another domain that will see a lot of research is privacy preservation using AI. Researchers are already working on ways to pre-process the data on the mobile device itself to ensure that only the extracted information is passed on to search engines and social media platforms and the privacy information of the users stays intact.
(You can watch the full interview at https://www.youtube.com/watch?v=pc9vsHlLhIg )