NIGRAANI: Voice-Verified Public Works Inspector

Public-infrastructure inspections are often scattered across paper checklists, WhatsApp photos, and reports that arrive days late. Supervisors can approve the next payment milestone without a clear trail of what was actually seen on site. NIGRAANI is a voice-first field inspection command center for public works. An engineer walks the site and speaks naturally, for example: "The final 120 metres of drainage are not built, and two areas have no safety barriers." NIGRAANI turns that speech into structured observations, compares them with the approved plan and the contractor's reported progress, and calculates a Truth Gap Score. The score is not an AI opinion. It is a deterministic, weighted-milestone calculation, so a supervisor can check the math. In our demo, the contractor reports 65% progress, the inspector's observations give 48%, and the gap is 17 points, flagged as review required. Trust is built into the workflow. The agent never accuses anyone, never blocks or freezes payments, and never submits anything on its own. If asked to block a payment, it refuses and offers to stage a supervisor review instead. Before anything is submitted, it reads back a summary, and the inspector confirms by voice and again by button. Every step appears in a live evidence timeline and audit trace. Each scenario is a data file, so a new project needs no new code. Four demo scenarios cover a progress gap, a quality issue, a clean inspection, and a contradiction. A printable supervisor report is generated after confirmation. All data is demo data and not an official government record.