SarkarSathi
AI co-pilot for municipal accountability in India.
Civic complaints in India vanish into systems that were never designed to respond. Citizens know what's broken in their ward (potholes, water supply failures, illegal construction) but translating lived experience into something a government portal will actually process is a skill most people don't have, and shouldn't need.
The problem isn't awareness. It's structured escalation.
SarkarSathi started as a complaint co-pilot: the AI helps citizens frame their issue clearly, identifies the right authority, and routes the complaint with the language and documentation that actually gets responses.
For India Innovates 2026, it evolved into full governance intelligence for elected representatives: commitment tracking across campaign promises, complaint clustering by district and severity, and an agentic advisor that surfaces what needs attention before it becomes a crisis.
Built this as the sole technical member on a 5-person remote team. During board exams.
Conversational intake via chat. OCR for photo evidence. Structured output that matches government portal formats.
FastAPI + vector database groups similar complaints across districts. A hundred separate "no water" complaints surface as one data point.
Representatives see a real-time dashboard: commitments made vs. kept, complaint surge detection, agentic recommendations on where to direct resources.
Gemini AI reads the complaint corpus and generates specific, actionable briefings. Not summaries, but decisions that can be acted on immediately.
