KaiSai

Attune is the Agentic Attention Router for live speakers. Every presenter—whether a teacher, pastor, or founder—is blind to the quiet confusion in the room. Questions get lost in chat, students are too shy to raise their hands, and the speaker has no idea which concepts actually landed. Attune solves this by listening to live audio, building a real-time Causal Graph of the concepts being taught, and using Meta's Muse Spark Open model to route audience questions by root cause. When a question arrives, the AI traces it through the graph, pulses the misunderstood concept in red on the speaker's dashboard, and clusters similar questions into a single actionable signal. Key features built during the hackathon: Live Causal Graph — a visual knowledge map that grows as the speaker talks. Anonymous QR Attendance — no audience audio is recorded, making it privacy-first by design. Agentic Question Clustering — groups duplicate questions and identifies the root cause of confusion. Low-Confidence Guardrail — if the AI is unsure, it defers the question back to the speaker with an amber warning, keeping the human in authority. Post-Session Report — after the talk, a batch report is generated on Modal's serverless GPUs, summarizing what the audience struggled with most. Why it matters: Attune gives speakers a sixth sense. No question is lost. No audience is left behind.