Manipal Prolearn Interviews Veteran Data Scientist at Flipkart – Ravindra Babu Tallamraju
Ravindra Babu Tallamraju
Principal Data Scientist – Flipkart
Veteran Data Scientist with more than 35+ years of experience
Previous Experiences – Infosys and ISRO
Ph.D. (Computer Science, Pattern Recognition, Data Mining, Large Data Clustering, Data Compaction) from Indian Institute of Science, Bangalore (IISc)
M.Sc Engineering (Computer Science, Pattern Recognition and Data Mining) from Indian Institute of Science, Bangalore (IISc)
M.Sc Statistics from Osmania University
Mathematical Modelling, Machine Learning and Data Mining
Manipal Prolearn interviewed Mr. Ravindra Babu Tallamraju to understand more about his journey of becoming a veteran data scientist.
Manipal Prolearn: Please give us your 30 second bio?
Ravindra Babu: I started working with ISRO and worked for about 24 years there. My work revolved all around Mathematics there. Those days a lot of importance was given to data mining. I simultaneously registered at the Indian Institute of Science, Bangalore (IISC). I pursued my M.Sc. in Engineering and also completed my Ph.D. there.
Subsequently, I left ISRO and joined Infosys, where I joined as the Principal Researcher. I was mainly working on the face recognition system there, and it was the first time I was getting exposure in this area. From the last two years I have been working with Flipkart as the Principal Data Scientist. That’s how my journey has been so far.
Manipal Prolearn: How did you get interested in field of data science?
Ravindra Babu: Data science is broadly about looking for data insights. So in that sense it is not significantly different from Statistics, which was the subject of my post-graduation. Data Science involves looking into the data and identifying relevant information, which can be used for business advantage. So it has always been there, maybe in a smaller magnitude, maybe for solving smaller problems with smaller sets of data.
When I started out in this field, the term ‘data science’ was not even coined back then, but everything that I studied is now certainly a part of what we call Data Science today. For me it still is broadly mathematical modelling of any given problem, finding insights and determining their validation, which I have been doing for last few years. So the names may be different – statistics, data mining, machine learning or data science, but the nature of work broadly remains the same.
Manipal Prolearn: Tell us about any interesting project that you worked on in ISRO?
Ravindra Babu: I have worked on several spacecraft missions in ISRO, and most of them were Indian Remote Sensing missions. I have also worked on the first Indian Operations Spacecraft project. Let me explain, typically, the spacecraft has an on-board camera, which records everything on the ground. My work involved predicting the spacecraft orbit and computing a host of derived information including camera foot prints. It was very interesting.
I have worked on several projects at ISRO and each had its own challenges, specifications, expectations. The challenges I faced have been different in nature – some of them have been minor and some path-breaking, based on which I had to form relevant models.
Manipal Prolearn: How has the field of data science evolved and why do you think it’s the next big-thing?
Ravindra Babu: As I said earlier, typically it all starts with data analysis. The amount of data that is now being generated is gigantic. For instance, you have billions of Twitter messages and you want to determine what is the broad topic being discussed in these messages. In fact, as a data scientist you may want to find out what’s the most trending topic on Twitter right now. For example, say I run a company and the trending topic is something that negatively affects my business. So, in this case I have to take action quickly and try to send a rejoinder for damage control.
Or a simpler example would be that of e-commerce companies trying to understand their user sentiment, preferences and how they can offer better services/products to their customers. All of this needs data to come to solid conclusions. Companies today are trying to save every data point available to them so that if not immediately, the information can be used later. So this attitude makes it vital to make optimum use of all the data you have and data science helps in doing that. So clearly this is just the beginning of data science.
Manipal Prolearn: How did you decide to join Flipkart?
Ravindra Babu: Honestly, I didn’t really choose it that way. In ISRO I had been working on various spacecraft projects for a long time and over the years I felt the need to work on something from the scratch, you know, to do things differently, and to learn new processes. At Infosys I got the opportunity to work on the face recognition system, which was a fairly new experience for me and it quenched my thirst to learn something new. After Infosys I was looking to work in a business related to manufacturing so that I have enough challenges to solve. That’s when Flipkart's then CTO convinced me that I will have enough challenges to keep me busy in my next stint if I join them.
Manipal Prolearn: Tell us about your typical day at work?
Ravindra Babu: Like always I reach around 8-8:30 to office. My day typically starts with digging deep into the available data and trying to build models based on that. So this role may not change on a day to day basis, but it certainly varies from problem to problem.
A very common problem that most online retailers face is the automatic categorization of addresses. This problem clearly impacts their optimal delivery system for ordered shipments. Once, I studied the supply chain I understood there are a lot of problems involved. For instance, even though Bangalore is a cosmopolitan city, I noticed that people from the same household did not necessarily mention their address in the same manner. In total there can be about 1000 ways people can write Bangalore with various abbreviations, spellings and the challenge is automatic categorization of these variations. This is a perfect problem to develop a model and I grabbed this opportunity as a data scientist.
The activity involves text processing, address pre-processing, clustering, classification using ensemble of classifiers, efficient ways to deal with large dataset with high dimensionality and increasing labelled dataset using semi-supervised classification.
Once the models are built, you have to analyse if they are working well, if they are offering the required solution, their accuracy percentage and so on. My work typically revolves around this.
Manipal Prolearn: Can you share some interesting data science trends with us that you have experienced while working in the e-commerce industry?
Ravindra Babu: Yes, a lot of my colleagues work on several data mining, instances which can be categorized as interesting and exciting. So they are constantly working on finding if the reviews uploaded on the ecommerce site are genuine or fake ones. It is important for an online retailer to judge the authenticity of the users uploading reviews on the site and ensuring that the reviews are not fake. Another aspect is determining if the reviews are actionable or not. Say the customer is not happy with the product, can we offer him anything to improve his experience with us? A lot of models are being worked on to automate these processes.
Manipal Prolearn: Which data scientist do you admire the most and why?
Ravindra Babu: I think it would be unfair to name one or two. Some of them are celebrated names, some are not but all of them are doing a commendable job. Take any company – Google, LinkedIn, Facebook, all of them have an array of data scientists working for them.
So, every time when these company-platforms offer you a quick solution, it is actually the effort of the data scientist behind the solution. So it won’t be wrong to say that they are making your life easier and I respect each and every data scientist who puts in the effort to make the world a better place.
Manipal Prolearn: Given an opportunity would you like to teach data science students?
Ravindra Babu: I have already done that actually! At Infosys, I was teaching data mining as a part of the company’s internal certification. I am also frequently invited to take guest lectures at colleges and educate students about the exciting field of data science.
Manipal Prolearn: Can you tell us some books that budding data scientists should read?
Ravindra Babu: I would say books that cover topics like statistics and probability, data aggregation, optimization, machine learning. Apart from that budding data scientists should also learn how to play with the data available to them. Books on R and Python are certainly helpful.
Manipal Prolearn: Tell us about a typical data scientist hobby that you have?
Ravindra Babu: I used to do cartooning earlier. However, I find my work so exciting that even when I am not at work I like to keep playing with data and developing models. I think that’s how data scientists should be!
Manipal Prolearn: What advice would you like to give prospective data science students?
Ravindra Babu: I would say there are no shortcuts in this field. You may read a book on Python, but your fundamentals need to be strong. Our work involves dealing with a lot of data. The right way to approach each issue only comes with a lot of formal reading and training. It is important to spend some time on the data available to you. Prepare the data, figure out the challenges. Analyse what is the ultimate goal and use the methods that are available with you to solve a challenge. But you should not compromise on the basics. If you do, you will end up with unclean data and ultimately an inefficient model. You need to have that excitement within you to work on new and challenging data sets. If you don't get excited at the mention of data, you certainly aren’t cut out for this field!
Manipal Prolearn thanks Ravindra Babu for the time and effort he put in with us. I am sure he is an inspiration to a lot of budding data scientists.
Data Mining Techniques Data Science Field Learn Data Science