Webpage
FIRST HACKATHON
designed a multi-agent system in Microsoft Foundry that goes beyond simple keyword filtering. Three detection agents run concurrently to flag different risk categories (harassment, threats, self-harm signals), and a reasoning-focused orchestrator agent weighs the context using gpt-6 model for heavy weight critical thinking to differentiate between actual harm and sarcasm, jokes between friends, or reclaimed language.