The Supplemental Nutrition Assistance Program (SNAP) is a vital lifeline for low-income families, yet structural barriers leave approximately 66% of eligible households in our dataset without aid. This project identifies the "SNAP Gap"—the highly vulnerable, legally eligible households that fall through the cracks. Using 2023 American Community Survey (ACS) data for MD, VA, and DC, we built a machine learning classification model (LightGBM) to predict SNAP non-participation purely through observable socioeconomic, demographic, and technological proxies. By identifying key invisible barriers like language isolation and legacy assets, this predictive tool empowers policymakers and community organizers to conduct highly targeted, data-driven outreach, bridging the gap between eligibility and actual enrollment.