Votify

Backend architecture overview
Landing page
Full dashboard
Stock market chart showing industry performance after policy implementation
Promise showing predicted vs actual industry impacts with credibility assessment
Promise detail modal showing evidence, sources, and multi-AI bias detection results

Inspiration

Misinformation about political promises is everywhere. Voters struggle to separate campaign rhetoric from actual policy action, often relying on biased media or their own faulty memory. We wanted to build a system that cuts through the noise with verifiable facts - a tool that holds politicians accountable using their own government records, not opinions or spin.

What it does

Votify automatically verifies political promises against official government records using multi-AI validation. It discovers promises from speeches and debates, cross-references them with legislative votes and executive orders from Congress.gov and WhiteHouse.gov, and determines if promises were kept, broken, or partially fulfilled. The platform includes stock market impact analysis showing real economic effects of policies, with every claim backed by verifiable government sources.

How we built it

Frontend: Vanilla JavaScript with Tailwind CSS for a responsive dashboard Backend: Node.js/Express API handling multi-service orchestration AI Pipeline: Perplexity for promise discovery, Google Gemini for legislative analysis, Cloudflare Workers AI (Llama 3.1) for bias detection Validation Layer: Python scripts with automated quality control and multi-AI consensus checking Data Integration: Yahoo Finance API for market impact, government APIs for legislative records Architecture: Discovery → Analysis → Multi-AI Validation → Bias Check → User Dashboard

Challenges we ran into

The biggest challenge was preventing AI hallucinations. Early versions would confidently cite non-existent bills or misattribute votes. We solved this by implementing multi-AI consensus - every Gemini analysis is cross-validated by Llama 3.1, and disagreements trigger automatic reloops until we get consistent, verifiable results. We also built strict source validation requiring direct government URLs, not just AI claims about sources.

Accomplishments that we're proud of

Built a working multi-AI validation system that actually catches hallucinations and bias Verified 40+ promises with 100% government-sourced evidence Created an automated bias detection pipeline that flags inconsistencies before they reach users Integrated real stock market data proving policy impact, not just political rhetoric Delivered a functional full-stack application in hackathon timeframe

What we learned

AI cannot be trusted blindly - even sophisticated models hallucinate confidently. Multi-model validation isn't just nice to have, it's essential for any factual verification system. We also learned that building civic tech requires balancing technical complexity with user trust - every claim needs a verifiable source link, not just an AI's word.

What's next for Votify

Scale to thousands of promises across federal, state, and local politicians Add real-time promise tracking as new legislation is introduced Implement user submissions with community verification Build browser extension for instant fact-checking while reading news Partner with civic organizations for wider voter education impact Add multilingual support for non-English speaking voters