Utkarsh Shukla is a Data Scientist at Blinkit. He joined us in July 2021 and has helped different teams make data-driven decisions since then. He is currently working on predicting item demand at our platform and has worked towards improving our predicted ETAs. His contributions include enhancing the GIS capabilities of the Data Team.
- Tell us about your background and your journey in Blinkit so far.
People who I work with would rather know me as UT. Hailing from Lucknow, I follow many sports (especially F1), but I hardly play any. I do love going for runs - helps in keeping the mind sane. Before joining Blinkit, I had worked with different startups, wearing many hats – solving engineering and data-specific problems. Eventually, I began my search towards finding a place where I could dive into data, work on building systems that drive business solutions and crib about data problems. I am fortunate enough to have gotten exactly that at Blinkit. I feel even before getting deep into data science; one needs to fall in love with data along with all the problems that come with it (believe me, there are a lot!). I am glad I have colleagues that share the same love for data.
Fun Fact: the place I used to work at previously had collaborated with Blinkit over data-related problems, and I was one of the team members responsible for the data preparation. So you can say I had been contributing to Blinkit’s data team even before I was a part of it.
- Why did you choose to work in tech? What drove you?
I seemed to have had an interest in computer programming from an early age. Credit goes to my school, where I got introduced to programming in the 6th grade. In college, I fondly remember my first encounter with a senior where he typed some code on an editor, and a dialog box appears asking for an email and password on a prototype of my college website. What fascinated me the most at that moment, was that I can execute and bring to life any idea/thought and present it to the world quickly.
That's great freedom and power to have. Getting into Data-Science specifically was the result of me trying to constantly find and solve different problems agnostic of the field. Computer Science and Mathematics are just tools that I have in my arsenal to tackle these problems. This ambition and my inherent interest in Math and computers led me to pursue DS.
- What excites you the most about your field?
Having the ability to solve problems and make decisions, not just through hunches or intuitions that one develops over time, but through well-formulated and explainable systems that validate/invalidate those hunches, is what keeps me in awe and excites me. Data Science has the added advantage of being accessible in terms of knowledge and is implementable with the least barriers.
- While solving a problem, what is the principle you always adhere to?
The first and foremost is to not blindly trust any data. Working with data is cumbersome, especially because faulty data not only affects the current process but has a huge impact on the downstream tasks as well. After working on many problems and dealing with different types of data, the only thing that keeps getting reinforced is that one needs to vet the data with a fine-tooth comb for the given problem at hand. Even before starting to find the solution with the given data, identify some data-specific and business-specific metrics that would validate its sanity. The second would be not to strive to get the best result on the first go. Doing that would quickly turn the project from a business problem to a research project. Always reach an acceptable baseline and iterate on it quickly using intuitions that you develop over time.
- What are the immediate challenges we are trying to solve in the e-commerce space?
Inventory planning and demand predictions are one of the beasts that every e-commerce tries to tame. Solving this helps in reducing dump-related cost burns, and having lower items in an out-of-stock state.
By providing a single, wide experience, personalization aims to generate individualized experiences for Blinkit users. This reduces the efforts customer put in to search for things they might need to buy
- What are your views on the future of technology in your domain?
I think with the increased interest in the field over the last few years, coupled with the fact that Data Science is now crossing many engineering/computing hurdles, data-driven solutions will become the common norm to solve problems, not only at an enterprise but also at a global level.
- What do you love the most about your role at Blinkit?
The execution autonomy and the push towards taking ownership of the problem is something I cherish. Having the trust that people have your back if things go south is an added bonus. Talking specifically with respect to my field, the sheer amount of data that the company has access to enables us to explore solutions for different problems. The best part is we get to deal with data that comes from all the different domains, like GIS data from delivery partner trajectory, text data from user addresses, time-series data from item sales, and many more.
- What would you say to a youngster interested in pursuing data science?
- To not just focus on getting theoretical knowledge in the field (which is important but not sufficient) - Working on multiple projects is a significant component in gaining the expertise needed to solve different data science problems.
- Fast iterations over finding one perfect solution. The complexity of the problem statements and intricacies of the data generally don’t allow us to find the best solution in the first attempt and the problem statement turns into a research project. It's always better to create an acceptable baseline on which iterations are made till we get results that solve the business problem.
- Tracking the right metrics - There are always standard metrics that are used to evaluate the performance of the system developed, but along with that the person should pay heavy attention to see if the system is solving the business problem and improving the related KPIs.
- Try not to learn everything - The Data Science field is very vast and any attempt to understand everything would just be overwhelming. The aim is to have clear basics and understand the intricacies of different methods and then deep dive into the system when implementing them.
- How do you strike the life-work balance?
I think this is something I am bad at. I get immersed in a problem and then the time devoted to solving it is unaccounted for. What I actually do is take regular time offs for two to three days, where I completely keep away from all tech to reset.
Thanks for reading this blog. Say hello on Twitter or follow us on LinkedIn.