Applying Machine Learning to Build a More Resilient World
A Q&A with Mohit Rikhy, Machine Learning Data Scientist at One Concern
Modeling the impacts of disasters is a highly complex, data-intensive process. At One Concern, we believe that this process requires constant innovation and application of new and untraditional technologies — the most prominent being machine learning and AI. From helping us resolve missing and incomplete data sets to enhancing the predictive and scalable capabilities of our models, our use of machine learning is essential to realizing our mission of building global resilience.
Mohit Rikhy is our machine learning data scientist, who helps us deploy ML in our models and conducts research on the best ways to apply this technology to our area of the disaster science field. We sat down for a remote chat about his favorite project at One Concern, our unique application of ML, and where he got his inspiration to pursue data science as a career.
What has been your favorite project since starting at One Concern?
My favorite project has been building the machine learning (ML) model for predicting the degree of damage a building would face (no damage, moderate damage, or extreme damage) after an earthquake for locations in Japan.
It gave me a ton of exposure to the intersection of machine learning, data science and structural engineering. It was interesting to tackle the challenges of having multicollinear features and the imbalance class problem to develop a robust model, and ensuring it is generalizable to different geographies and unseen events. Apart from that, I had to collaborate with domain experts in several different fields to build this model. During this process, I came across challenges which I hadn’t experienced while designing ML models for other industries in the past, which really helped me expand my horizon and skill set. Moreover, it was extremely satisfying to be able to apply my ML skill set for such a noble cause of doing better for humanity.
What are some of the challenges of employing machine learning in our product? Any hard problems in particular that you’re focusing on at the moment?
The foundation of machine learning is data — machine learning models rely heavily on it. The better the quality of data and the more data we have, the better we expect our models to perform. The biggest challenge of employing ML in our product is dealing with the scarcity and disparity of data. Earthquakes are extremely rare in nature, and being an extremely niche field, there isn’t much data around which we can find easily for our use cases. When we do, we mostly get unstructured datasets at very different levels of resolution and quality. Processing and combining them, while trying to reduce the uncertainty around them in order to be used by ML models, is the biggest challenge we face.
Can you talk a little bit about why we use machine learning to help us model the impacts of disasters? How does it help us understand disasters in new ways?
Machine learning is essentially learning from what has happened in the past to try to predict what is most likely to happen in the future.
Natural disasters have a lot of randomness associated with them. Currently, there are sophisticated physics models to model the impact of disasters. However, they are known to have large uncertainty; and for earthquakes, they typically depend on only a single shaking feature. An ML model, on the other hand, can take multiple features into account and can be tuned using historical damage data to reduce uncertainty.
By developing ML models, our goal is to reduce the randomness involved in natural disasters by learning from past events, in order to make more accurate predictions of disaster impacts. Using ML can help us overcome the shortcomings that current state-of-the-art models face.
These models can then help us run different scenarios by altering earthquake and building parameters to see the different possibilities of damage, helping us understand building characteristic behaviors in different ways and measure the vulnerability of buildings and cities.
In a few words, how would you describe your team?
I would describe my team as smart, positive, driven, helpful and empathetic. It is indeed a great feeling to be surrounded by extremely positive and smart people whom you can learn so much from. It amazes me how much I learn from them everyday, whether it is on the professional or personal front. Everyone on the team is super eager to help out and unblock one another, whenever it’s needed, which I think makes this team so unique and enjoyable to work with.
Now, a few fun questions about yourself…
Do you have a favorite quote?
Yes, absolutely. I love thought provoking philosophical quotes. My favorite one is:
“In the end, only three things matter: how much you loved, how gently you lived, and how gracefully you let go of things not meant for you.” — Gautama Buddha.
What did you want to be growing up?
Interestingly, I never really had one profession in mind while I was growing up. My interests were always evolving. I believe my inherently curious nature always wanted to try out different things. Sometimes I wanted to become a pilot (like most young boys), sometimes a doctor, and sometimes even an actor (even though it is hard to believe). But in hindsight, I feel it all makes sense since I feel I have put my curiosity to the best use with a career in data science. For me, data science is nothing but being curious about your data and trying to make the most of it with constant experimentation, with the goal of creating a positive impact. And I am trying my best to achieve that.
Finally, what does resilience mean to you?
For me, resilience is adaptability. Change is the only constant, and being able to adapt to changing conditions in one’s life seamlessly is how I define resilience. Everyone faces turbulence in different aspects of their lives, and coping with any of these can be exhausting. I strongly believe that one should take any possible help or support (emotional, informational, instrumental etc.) that can make adapting to these changes more comfortable.
The previous 9 months have shown us how severely the world can change all of a sudden. The world has changed in the way it operated, to the extent that we have forgotten what the norm was. These last few months have made me realize that resilience starts with you. If you are more resilient, fostering resilience around you is easier. And the first step to that is understanding and accepting that one would need to adapt and change as time passes and circumstances change, and embrace any change with the best spirit and belief that it would be for the better in the long run.
Interested in building planetary-scale resilience? We’d love to hear from you! Check out our open positions or reach out to careers@oneconcern.com.
Related stories:
Solutions for a Changing World
Building Resilient Teams