DowntownDry | AI Venue Verification System

Project Overview

DowntownDry is a Nationwide directory of venues that offer Non-alcoholic beverages, kava, THC, and other alternatives to alcohol. It utilized an AI-powered venue verification system that automates the process of validating and verifying business information at scale. The system integrates multiple Google APIs with Gemini AI to achieve exceptional accuracy while dramatically reducing manual review requirements.

The Challenge

Manual venue verification is time-consuming and error-prone. Businesses need to verify hundreds of venues across multiple data sources, cross-reference information, and ensure accuracy before publishing. This process traditionally required significant human review time.

The Solution

I built an intelligent verification automation system that combines:

• Google APIs Integration: Places, Maps, and Business APIs for comprehensive data gathering
• Gemini AI Processing: Automated analysis and verification of business data
• Smart Validation Logic: Cross-referencing multiple data sources for accuracy
• Human-in-the-Loop Design: Flagging edge cases for human review while automating clear-cut decisions

Key Results

• 97% verification accuracy - Matching or exceeding human reviewer accuracy
• 90%+ reduction in manual review time - Freeing up team members for higher-value work
• Scalable processing - Handles hundreds of venues automatically
• Consistent quality - Eliminates human fatigue and inconsistency issues

Technical Implementation

The system was built using Astro.js for the frontend, with a robust backend using Supabase that orchestrates calls to multiple Google APIs. Gemini AI handles the intelligent decision-making, analyzing business data across multiple dimensions to determine verification status.

Responsibilities

• Architected the full verification automation pipeline
• Integrated multiple Google APIs (Places, Maps, Business)
• Implemented Gemini AI for intelligent verification decisions
• Designed the human-in-the-loop review system for edge cases
• Optimized for processing speed and accuracy