Decisions and Deployment: The Five-Year SAHELI Project (2020-2025) on Restless Multi-Armed Bandits for Improving Maternal and Child Health

The SAHELI project (likely associated with the Tambe Lab at Harvard/Google Research and partner ARMMAN) represents a rare 'high-defensibility' research project. While the core algorithms (RMAB) are known, the 5-year longitudinal deployment data and the integration with specific NGO workflows (like ARMMAN in India) create a massive barrier to entry. The 'moat' here is not just code—which has 5 forks but 0 stars (typical for a just-released research artifact)—but the domain-specific reward modeling and the validation of AI-driven interventions in low-resource settings. Frontier labs are unlikely to compete directly here as this requires intensive 'on-the-ground' partnership and social-impact focus that doesn't align with their high-margin compute-scaling models. The risk of platform domination is low because big tech (Google/Microsoft) typically funds this work as CSR/AI for Social Good rather than productizing it to compete. The displacement horizon is long because the project is uniquely grounded in a 2020-2025 multi-year validation cycle that cannot be simulated or shortcut with larger LLMs.