Team Daksh(ProofCam)

ProofCam — Giving AI the ability to verify work in the physical world. The world is building increasingly sophisticated digital infrastructure, but the physical systems powering it are still verified through fragmented photographs, manual checklists, and human interpretation. From hyperscale data centers and telecommunications networks to industrial manufacturing, renewable energy infrastructure, healthcare equipment, and smart buildings, organizations depend on thousands of physical installations being completed correctly. Yet a photograph of completed work is not necessarily proof that the work was done correctly. ProofCam addresses this fundamental gap. We built an AI-powered visual inspection agent that transforms an ordinary smartphone into an intelligent evidence-gathering system. Instead of simply asking AI, "What is in this picture?", ProofCam asks a more consequential question: "What evidence is still missing before we can establish that this installation meets its requirements?" A technician selects a predefined installation specification and photographs the physical setup. ProofCam uses Meta Muse Spark 1.3 to understand the scene and reason about missing evidence, Meta SAM 3.1 to segment and spatially ground objects, and a deterministic verification engine to evaluate the installation against its requirements. When evidence is incomplete, the system doesn't guess. It guides the operator toward the next useful camera angle. When it finds a supported contradiction, it can report a violation. When the photographs cannot establish a required fact, it explicitly returns INSUFFICIENT_EVIDENCE rather than manufacturing confidence. Every inspection produces a traceable record of observations, visual evidence, unresolved claims, decisions, and model activity. Our working prototype demonstrates this through multi-device, multi-connection hardware inspections, including routers, networking equipment, gaming consoles, and displays. But the opportunity extends far beyond cables. We envision ProofCam becoming a reusable verification layer for physical operations: an intelligence system that organizations can eventually connect to their existing installation standards, deploy across distributed workforces, and use to review physical work consistently. The next frontier of AI is not just understanding the digital world. It is helping us establish what actually happened in the physical one. ProofCam is our first step toward that future.