Skip to Main Content

SpillSight

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

SpillSight — Demo video

SpillSight helps teams respond to oil spills faster. I built it for the Strait of Hormuz, where about a fifth of the world’s oil passes, after a damaged tanker there spilled oil that reached the UAE coast where I live, close to a plant that turns seawater into drinking water. The hardest part of a spill is the first few hours. Today, experts work out by hand where the oil will drift, and that takes time. Crews often go to where the oil was, not where it is going. Barriers placed at the wrong angle let oil slip under them, beaches and water plants are warned late, and when the wind changes there is no time to redo the plan. SpillSight puts the whole response on one 3D map. First, you load a satellite, aircraft or drone image, and Meta’s SAM 3.1 outlines the oil in about two seconds and places it on the map. Next, the system releases three hundred virtual drops of oil and moves them with the wind and the ocean current for three days, stopping them where they reach land. This shows where the oil will come ashore and how many hours each beach, reef, village or water intake has left. Then SpillSight looks at every boat, barrier, skimmer and drone available, where each one is and how fast it can get there, and Meta’s Muse Wonder turns all of that into a plan. Each asset gets a job, an arrival time and a short reason. Two boats tow a U-shaped barrier around the spill, angled so the current cannot push oil under it, and skimmers wait at the closed end to pump the oil out. More barriers protect the reefs, and nearby shores are warned hours before the oil arrives. Nothing happens until a person approves the plan. After that, the boats leave port on the map, the barrier closes around the spill, and each task moves across a board as crews arrive. When a new image comes in, the whole process can run again in under a minute.

Team