Krishna Sai Vootla has worked as a business analyst before becoming a data scientist. He graduated in electrical engineering from the Indian Institute of Technology, Gandhinagar and has worked for companies like Tredence.inc, JPMorgan Chase and co., etc.
“When I joined my current company, the data warehouse was in a really bad stage. Data from all sources were not integrated. Moreover, the reporting was not automated. I did a lot of Data Engineering and ETL. It was a fulfilling experience to implement a full warehouse, end-to-end ( from API integrations, script automation, ETL, visualisations & ML code deployments),” said Krishna.
Sign up for your weekly dose of what's up in emerging technology.
Analytics India Magazine got in touch with Krishna to understand his data science journey.
AIM: Is a degree in data science enough to get a job?
Krishna Sai Vootla: A degree is not enough. I always believe in “slow and steady wins the race”. It is helpful if a person starts early. Institutions like IIT Hyderabad have a bachelor’s degree in AI. Though data science and ML have existed as a science for a long time, they have been put to good use in the past two to three decades. I believe 90% of the research level development took place in the last decade. Even if someone starts late, if they are consistent, they can become an expert in a year or two.
AIM: Tell us about your data science journey.
Krishna Sai Vootla: To be frank, I had no clue about data science before 2016. I joined a team working on a project for the Intel Embedded Systems Design Contest in 2016. We worked on a prototype to help medical emergency response teams make quick and accurate diagnoses. I did not have any particular interest in programming or computer science until this point. But, I enjoyed developing code bits for my project. Having gotten the taste of it, I wanted more, so I started taking up programming tasks. I started learning C++ in my spare time and was able to pick it up quickly. In just a couple of weeks, I learnt the basics and contributed to the project. I taught myself programming, OOPS, Python, basic machine learning concepts, SQL etc with the help of MOOCs and academic blogs. Andrew Ng‘s course is the best introduction to data science and machine learning. In his course, he talks about many topics and how transforming technology can be useful in various industries. I started exploring a lot of things and realised the tipping point we are in, and understood the potential of data. Decision making in the past was based on a mix of domain knowledge and approximated estimates. Today, businesses make data-driven and educated decisions. I have always been a numbers guy. Maths and Physics were my favourite subjects in high school. Conveniently, my work needs me to be on top of numbers doing analytics to make data-driven decisions for business growth.
AIM: What are the interesting projects you have worked on while in college?
Krishna Sai Vootla: Multimodal CNNs was a research project about using multiple streams of image input to see if it helps the model understand better. We found a dataset with forest scenes and trained a convolutional neural network to segment scenes of these images. Further, we modified the architecture to add multiple visual modalities apart from RGB (like images from a depth camera). We found that adding multiple modalities of the same scene improved the segmentation accuracy over just RGB images. Additional modalities proved to act as a new pair of eyes for the model to learn from. It is difficult to comprehend whether the neural networks, being a black box, learnt the right criteria or not. For example, the model could associate anything with blue as the sky if our training data has more images with a blue sky and few images with a blue lake. But with additional modalities like depth images, the model can differentiate these two; the depth value of a lake is always less while that of the sky is very high.
Telepresence Robot: We used computer vision to decode human movements and worked on translating those motions directly onto a robot.
Smart Medical networks: We worked on a prototype that could help medical emergency response teams in making quick and accurate diagnoses.
AIM: How did you land your first job at Capgemini?
Krishna Sai Vootla: I got a campus placement in Capgemini and joined an internal team as an ETL developer. My work forced me to learn the most important and employable skill that I would use every day, SQL. I learnt SQL in college, but working as a part of an ETL team pushed me to learn and re-learn all the nitty-gritty.
Then, I moved to Tredence, a start-up at the time. While I taught myself ML and DL in college, I have learned analytics skills like Excel (everyone underestimates this, but it is really powerful for starters to learn, although not scalable), Tableau, Statistics, R Programming at Tredence. I worked on a customer segmentation project for Walmart. We built a model to segment the 100 million customer base into various segments for the effective allocation of marketing resources and the maximisation of cross and up-selling opportunities. It gave me a lot of exposure and experience on how models are developed, deployed and how data science tools can help business growth.
AIM: Why did you choose to become a freelancer?
Krishna Sai Vootla: Currently, I work as a freelance Data scientist at Toptal Talent Network, and my current Client is Organifi LLC, a superfood company in San Diego, with whom I’ve been working for over a year. My key job function is to derive insights and build KPIs that help high-level data-driven decision making. Recently, I built a sentiment and opinion extraction tool. I have compared accuracy results from different combinations of features and models. The tool categorised the customer reviews and scored the sentiment values to identify pain points in specific products and customer service. Fixing these led to consistent growth in sales and a decrease in the number of negative reviews over the next few weeks.
I aim to contribute to open-source data science/ ML projects. I want to build tools, products for digital well-being. Also, I am working on a couple of personal projects. Recently, I deployed a code on my PC that records my emotions every minute and stores them in a text file. After a month, I will summarise how my emotions change at any point of the day.
AIM: How is the freelance scene for data science professionals?
Krishna Sai Vootla: There are a lot of opportunities outside India. I believe it’s the right time for Indian software engineers, data scientists and IT professionals to go global like other countries. The Covid pandemic definitely catalysed this scenario. There is quite a bit of taboo in India about not having a secure, permanent job. I, too, faced some resistance and questioned/ doubted myself multiple times in this journey. But my colleagues at Toptal gave me a lot of confidence. Now, I am at peace with the uncertainty of freelancing because I trust my skills and know there are a lot of businesses that require a good data scientist.