• Designed computer vision breast cancer detection model based on a case-based reasoning architecture in PyTorch.• Enhanced interpretability of the model by making it more sensitive to fine-grained features in images. Achieved SOTA AUC scores of 0.95.

• Developed Machine Learning models for earthquake early warning systems set to be installed in earthquake detection facilities in Nepal.•Implemented pipeline that inputs seismograph wave data, transforms the data to wave features, then uses the trained model to predict intensity of oncoming earthquakes to trigger alarms. Improved upon the SOTA R2 scores by 30%.

• Standardized and centralized health data from multiple sources into a single repository to be used by multiple institutions.• Analyzed Clinical Trial and survey data to generate insights and validate hypothesis on child health outcomes.• Secured $200,000 in funding from the Bill and Melinda Gates Foundation to lead a project on analyzing the impact of improved household income on child undernutrition in India.• Modelled child undernutrition in the country using Geospatial Models (Gaussian Process Regressions) using PyTorch. Identified three key regions with potential for 15-25% reduction in undernutrition through improved monthly income.• Publications:- Kishore S, Thomas T, Sachdev H, et al. Modeling the potential impacts of improved monthly income on child stuntingin India: a subnational geospatial perspective. BMJ Open 2022;12:e055098.- Shivakumar N, Kashyap S, Kishore S, et al. Protein-quality evaluation of complementary foods in Indian children. American Journal of Clinical Nutrition. 109:5. May 2019. Pages 1319–1327.

Lighthouse Datalab was a Data Science consulting company offering analytics solutions. During my internship, I consulted for a leading healthcare blog website.• Developed machine learning models to help a healthcare blog client perform demographic based ad targeting.• Trained tree-based ensemble models to classify user’s age and sex, improving AUC by 10%, ad click-through rate by 5%.