Midas Ledger

Midas Ledger Poster
Home Screen
Results

Inspiration

Millions of people miss out on government benefits simply because eligibility requirements are difficult to find, understand, and compare. Information is often scattered across websites, documents, and application portals, creating barriers for those who need support the most. We wanted to build a solution that makes welfare programs easier to navigate through a simple conversational experience.

What it does

Midas Ledger - Benefitor AI is an AI-powered benefits navigator that helps users discover government welfare programs they may qualify for. Users describe their circumstances in natural language, and the platform analyzes their information, matches them against relevant schemes, and provides explainable recommendations. Each result includes eligibility reasoning, required documents, and application guidance.

How we built it

The platform combines a modern web interface built with Next.js 16, React 19, and Tailwind CSS with a FastAPI backend. Gemini 3.5 Flash is used to extract structured information from user conversations, while a deterministic Python rule engine evaluates eligibility against predefined scheme criteria. This approach keeps recommendations transparent, reliable, and easy to explain. The platform currently supports eight languages and follows a privacy-first, zero-PII design.

Challenges we ran into

One of the biggest challenges was ensuring that AI did not become the decision-maker. Early approaches occasionally produced inaccurate eligibility conclusions. We solved this by separating language understanding from eligibility evaluation and moving all decision logic into deterministic code. Another challenge was converting complex welfare criteria into clear and consistent rules.

Accomplishments that we're proud of

We successfully built an explainable benefits navigator that combines conversational AI with transparent rule-based eligibility matching. We are particularly proud of the multilingual support, privacy-focused architecture, and a design that prevents hallucinated eligibility decisions while remaining easy for users to understand.

What we learned

This project taught us that trustworthy AI comes from thoughtful system design, not just powerful models. We learned the importance of keeping critical decisions transparent and auditable while using AI to improve accessibility and user experience.

What's next for Midas Ledger

We plan to expand coverage to include more national and regional programs, improve multilingual capabilities, and refine eligibility rules as policies evolve. Our long-term goal is to make benefit discovery faster, more accessible, and easier to understand for everyone.