CraveQuest — Crowdsourced Foodie Super App

The Challenge

Restaurant menu prices are constantly changing and rarely available online. Consumers can't compare prices or discover nearby deals, and restaurant data aggregators rely on outdated, manually entered information.

Solutions

• Built a 9-step AI processing pipeline: OCR (Cloud Vision) → Gemini 1.5 Flash structured extraction → multi-layer fraud detection → menu item upsert with fuzzy deduplication.
• Designed a full gamification layer with CQ Coins, bounties, leaderboards, a weekly lottery, and a 20-level XP progression system to drive user engagement.
• Implemented a geo-aware map interface with PostGIS queries, real-time bounty markers, and Elasticsearch-powered menu search with Turkish language support.
• Engineered a layered anti-fraud system: moiré pattern detection, physical document confidence scoring, receipt hash deduplication, GPS velocity checks, and an automatic shadow-ban engine.

Outcomes

• Full-stack platform at 86% completion across 6 development phases (126/147 tasks).
• Scalable GCP infrastructure via Terraform: Cloud Run, Cloud SQL (PostGIS), Memorystore (Redis), GCS, Elasticsearch.
• Production-ready Flutter app with Riverpod state management, secure token handling, and camera-only geofenced uploads.
• CI/CD pipelines via GitHub Actions with Workload Identity Federation for secure GCP deployments.

Stack