A4
Track A — Conversational & RAG-Powered Assistants

Government Scheme-Matching Assistant

Agents that answer questions by retrieving the right information and responding conversationally, in the right language.

Background

Agents that answer questions by retrieving the right information and responding conversationally, in the right language.

Objective

Develop an agent that collects a short profile from a user and identifies exactly which government schemes they are eligible for, explaining the reasoning behind each match.

Key Features

1. Conversational profile intake (5–8 questions) 2. Rule-based eligibility checking against curated schemes 3. Reasoning shown for each match ("why you qualify") 4. Near-miss detection — shows what would make someone eligible 5. Ranking of matched schemes by relevance/benefit

Expected Solution

A conversational intake form that adapts questions based on prior answers An eligibility-reasoning engine (rules + LLM explanation) A ranking and report-generation layer

Suggested Technology Stack

Backend: Python / FastAPI Reasoning: Rules engine (Python) + LLM for explanation Orchestration: LangGraph Data: Curated scheme dataset (MyScheme.gov.in, PMEGP, MUDRA, Stand-Up India, etc.)

Expected Outcomes

Increased scheme uptake among eligible but unaware citizens Reduced dependence on intermediaries/agents for scheme discovery A transparent, explainable eligibility-matching pattern reusable for other domains

Possible Use Cases

A first-time entrepreneur checking loan/subsidy eligibility A student checking scholarship eligibility An NGO field worker helping community members discover schemes

Evaluation Parameters

Correctness of eligibility logic Clarity of reasoning shown to the user Scheme coverage Overall UX