7 learnings from interviewing top leaders in machine learning and data science
By three methods we may learn wisdom: First, by reflection, which is noblest; Second, by imitation, which is easiest; and Third, by experience, which is bitterest.
— Confucius
We started DataHack Radio 3 months back. Some of you might ask — why another Podcast on Data Science? The answer is same as it was 5 years back when I started Analytics Vidhya :
Creating a podcast helps us learn from top brains in this domain and share it with the community. It helps us build a perspective, which might take years for us to build by ourself.
So, now that we have a few episodes out, it is a good time to reflect what I learned and share it with our community:
- Top leaders in industry come from varied background and promote diversity. Anthony Goldbloom and Emily Sands Glassberg were both Economists, Marios came from Finance and Accounting background, Carla Gentry studied mathematics. Some of the people did not even know programming when they fell in love with data science. Bottom line is that Data Science as a field, actually becomes better by getting people from different background to come together. Just because you are not a Computer Science graduate does not mean that you will not be able to become a data scientist.
- Top leaders are always learning new tools and techniques. I think the episode with Anthony stands out on this. Not only Kaggle is constantly enabling research through competitions, they are constantly evolving Kernels to bring in the latest developments for their community. Anthony describes how Kaggle enables people to create, test and popularize various tools and techniques and how XGBoost got popular on Kaggle.
- Knowing the tools and techniques is important, but that alone does not take you to the top. You need to understand the challenges in implementing machine learning in business. I love the way Emily described her projects and insights — how she defined the problem, the underlying data structures and what it enables Coursera to achieve.
- A lot of world problems still need data collection, data cleaning, data augmentation and making it ready for work. The episode with Dr. Avik Sarkar stands out here. Most of the problems his team is working on includes collecting data through various primary, secondary and tertiary sources. One such problem he is working on includes predictive yield management for farmers in India. His team is working on Satellite image data, weather data, pricing from various marketplaces and past yield data to help farmers assist in right time for growing their crops.
- Data privacy is top of the mind for these professionals. The discussion with Carla Gentry brings this out clearly. We talk about GDPR being one of the early changes in the way Privacy would evolve. She expects more such changes in coming years and asks data scientists to take the onus of making sure that we do the right data science for the customers, while respecting their privacy.
- Collaborative learning and building open source is the future of data science. Kaggle is clearly taking a lead here — open datasets, Kernels, discussion forums. But, this is one theme which runs common across these interviews — Kiran mentions his team building an open sources library for EDA, Tarry Singh and Marios talk about this in great detail. Tarry Singh enables people across the world with his workshops and Marios talks about how much he has benefitted from democratisation of data science.
- Competitions are not an end in themselves — but they still provide you an awesome playground to test your data science concepts and make you ready for accelerated learning. Anthony, Marios and Kiran all agreed that there are times when competitions might not reflect the challenges in real life.
These are the top learnings as per me. There is lot more in each of these episodes and discussions.
If you listened to any of these, what were your top learning? Do share it with us. Till then — Keep Learning
