Pre 2025
"Research is the distance between an idea and its realization."
- David Sarnoff
Machine Learning Methods for Enhanced Forecasting of Antiretroviral Therapy Demand in India
This project was part of the initial cohort of the Breakthrough Research Grants Program, a joint collaboration between the Gupta-Klinsky India Institute at Johns Hopkins University and Koita Centre for Digital Health- Ashoka University. This project focuses on building an AI-driven forecasting system to improve the prediction accuracy of Antiretroviral Therapy (ART) drug demand across India.
HIV treatment in India faces multiple operational challenges due to long procurement timelines, seasonal migration, dosage changes, and frequent medication substitutions. Developed in collaboration with the National AIDS Control Organisation (NACO), the project has two major phases:
- Engagement and System Integration: The initial effort was to engage NACO and build a high-security system capable of interfacing with their data infrastructure.
- AI-Powered Forecasting and Drug Regimen Modeling: This phase uses machine learning models to forecast drug demand two years ahead, accounting for variables such as seasonal migration, patient-level drug switching, pediatric transitions, and center-level anomalies.
The data provided by NACO was highly unstructured, often shared via inconsistent Excel files. An AI-adapter pipeline was developed to clean, parse, and process this messy data automatically. These predictions are made at the ART center level (700+ centers across the country), allowing for micro-level accuracy. The tool also integrates a user-friendly interface tailored for use by healthcare workers with limited technical training. The ultimate goal is to enhance forecasting precision, reduce over-purchasing, and ensure more efficient distribution of HIV medication nationwide.
