ElasticShelf
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

ElasticShelf is a headless dynamic-pricing and replenishment layer that sits between a grocer's systems and its electronic shelf labels (ESLs). The problem: grocers lose billions to perishable waste — one-third of all food is lost or wasted (FAO), 30–40% of the US food supply is never eaten (USDA), and US grocery retail alone generates 4.45M tons of surplus food (ReFED, 2023) — all on ~2% net margins (FMI), while shelf prices sit static. ElasticShelf closes that loop every simulated hour: it reads weather, events, and store hours, scores each SKU's demand situation, reprices perishables through learned policies with safety guardrails (never below cost, never above 1.5× base), pushes the new prices to the ESL wall with attached reasoning, and forecasts restock quantities the moment a SKU depletes. A paired-draw no-tool twin running alongside proves every dollar of gain comes from pricing in our demo run, +$552 net profit with waste cut from 137 to 46 units.