Anuradha Bulusu is an AI researcher and data scientist who is building predictive models to detect population health and vulnerabilities.

Can you share the most interesting part of your career journey?

The most interesting part of my career journey is seeing how I can apply my research, experimentation, and analytical reasoning skills across a wide variety of fields ranging from quantum modeling of electrons or nanomaterial design in my prior academic career to my current role at Deloitte as an AI researcher and data scientist, where I build predictive models for population health and vulnerabilities.

What excites you most about working with data and AI?

The opportunity to use data and building predictive models that are used by clients to solve real-world problems. My work focuses on building explainable models to predict risk of a variety of social vulnerabilities in populations, such as uninsurance risk, disease progression risk, food insecurity, etc. Every one of these models are used by numerous government and industry clients to identify people in need and perform targeted outreach.

There is no greater joy and satisfaction than being able to see the fruits of your work being used to make an impact in real time.

During the COVID pandemic, we worked with all 50 US states and even the US islands and territories to provide data-informed recommendations to identify at-risk populations for microplanning and outreach. There is no greater joy and satisfaction than being able to see the fruits of your work being used to make an impact in real time.

Describe an interesting project that you have worked on.

Building machine learning models to predict population vulnerabilities brings a unique set of challenges given the complexity of the problem and the accompanying data challenges.

My team and I are currently building models to predict the likelihood of an individual’s risk of having a variety of diseases such as diabetes, hypertension, opioid abuse, alcohol abuse, depression, etc. Given that medical claims data and the associated data collection is incomplete and imbalanced for a variety of reasons, we need to carefully design and conduct the appropriate experiments to identify and measure data and model bias across age, race, gender, and other demographic cohorts and come up with the appropriate mitigation techniques.

The outcomes of these models are being used to study disease progression across a longitudinal population cohort to identify strategies to minimize disease risk.

The learnings and client outcomes derived through this work have been incredibly enriching and rewarding.