Name: Swapnil Pote
Title: Principal Data Scientist
The leading AI man, Swapnil before being the team lead is the mentor, the friend, the philosopher who shares all his accolades with his team. For him, all the accomplishments are teamwork and rightly so.
A true tech genius who, besides heading the AI team at OGenie, also heads his own AI-based startup. But don’t for a minute think of him as the guy focused only on work, he pursues a lot of other passions as well.
I’m surrounded by data all day- collecting, segregating, training, re-training, building models, discussions- spending up to 80% of my time prepping the data. It’s just amazing to see how little pieces of data picked from here and there come together and define big traits. But as the day progresses so does the struggle, each day to get past the point where results become a constant.
The number one contributor to this struggle is data quality. When we are collecting data to train the model, the data is not diverse enough to represent the variety of edge cases. Because of this, edge cases that are not seen frequently may be underrepresented when training and testing so the model ends up not being great at finding these. A machine learning model is only as good as the data you are putting into it, and there is a lot that can go wrong.
Giving in so much time to a model and then realising that we missed out on a detail. That’s kind of a big low for us. But that’s the beauty of working for OGenie, we fail, we learn, we start again. We just can’t afford to waste time over our failure, so it’s not a failure but learning.
Building a great team!
Working on machine learning and data learning is interesting, but taxing as well. The challenges come and go, and believe me when I say we’ve faced a lot of setbacks, but even the biggest challenge can be overcome with a great set of passionate doers.
One of the major points of failure in AI projects is the jump from a working model in the lab to a working model in the system. Organisations have high hopes from AI models to improve automation, and once they implement they may find their model is not performing at the accuracy level desired.
They end up going back to the drawing board to iterate on the model and to their old way of doing things. Not only this becomes an issue for the project, but also an issue of perception for AI. Don’t rush into implementing the solution, let it take the time it requires, the data it needs and the iterations where needed.
To optimize a deployed solution, continuous learning is one of the most important aspects. The first step is having a method for continuous data collection. As the model is making predictions it’s important to be able to collect data where it may have struggled.
Periodic retraining of the model is also very important. This periodic retraining on new data helps to optimize performance by catching and training on those infrequent edge cases.
I’m connected with a lot of groups- developers, scientists, early adopters, tech enthusiasts and on a lot of forums and platforms- Quora, Medium, GitHub, Stackoverflow.
I read, I listen, I write, I share- let the knowledge flow.
Let’s get Personal
Right after my college. A bunch of friends working on transforming the world one step at a time.
My work yes, but not my voice. I’m a shy speaker. Also, an introvert. Everything I am or do is a part of my life. I don’t let anyone thing define me. But my consistency towards achieving something that can define me.
Very near yet very far. Every time I come closer to my goal, I see a lot of pathways to learning more and I seek them, so very near yet very far.