Distinguished Architect & Head of Strategic Machine Learning Enablement, PayPal.
What’s your background and how did you get into Data Science?
I am responsible for architecting high-performance AI systems with a deep understanding of underlying data in PayPal. Along with statistical modeling, I combine statistics and data science to provide powerful models making strategic decisions.
My passion for Data Science started at a very early stage when I was doing my double major in Statistics and Operational Engineering. My interest in the trio combination of Computer Science, Statistics and Math paved my path towards deep understanding of data and applying computation deriving valuable predictions.
What do you think about the current industry adoption of Data Science and Machine Learning?
Industry adoption of AI is very vague. You have to respect the fact that it is different for different people. A lot of people come in with previous knowledge of the industry that they worked in. You should always use your previous knowledge and carve out the usage of Machine Learning to improve your business.
What people call AI is all over the place. Some people term a simple rule engine as AI. But in healthcare, the level of AI adoption is increasing from automatic image recognition to patient health prediction. So the AI adoption varies based on industries.
What are the challenges in adopting AI?
Google Machine Learning can run complex TensorFlow but does not have a GUI. One have to start learning Python, Tensorflow, deployment strategies if they want to make use of their data and domain knowledge. This alienates people from stepping into AI. We don’t have anything in-between.
We are witnessing abundance in data. But are these data readily available? People who understand data are not the person solving the data problem. This disconnect is our biggest hurdle. For executives, AI is a BlackBox. Business Leaders often visualize AI as a magic box solving all their problems and if they do not get the result they anticipated, they are disappointed.
How can we improve AI adoption in any enterprise?
Fabric team structure is the way to enhance AI adoption in any enterprise. One centralized Data Science team should be governing the AI practice guiding data scientists across various sectors.
Educate all sectors of people about AI and its usage. Start with a simple use case, not one simple or not one large one, but start with 3 or 4 simple use cases. Some will work and some may not. Adopt the ones that show promising results. You should care about building effective models with deployment tightly aligned with model execution.
Machine Learning solves the hardcore part of the computation. But in order to build effective models, you need to understand the business domain and how data can be used to enhance business. For example, if your domain is biology, then apply Data Science in biology instead of optimizing neural networks to predict the next kind of enzyme. Use ML for the appropriate business use cases.
What is the future of AI?
AutoML will untimately increase the adoption of AI as anyone can use the automated machine learning models to engineer the data for powerful insights improving business.
It is a great first step trying to bridge the gap where domain experts can use their domain knowledge and data without being a technologist. Business can start employing AutoML and understand the business problem to get the feel of what is possible.
AutoML is democratizing AI and anyone can become a citizen Data Scientist
One final advice for all the budding Data Scientists
Find what you are interested in and stick to your gut. Join like-minded groups and enhance your domain knowledge. Come with an open mind. You cannot quench the Data Science quest in its entirety. Start small and grow from there!
Success will follow you when you stick to your gut and follow your passion