Nutrition tools are often fragmented.
Barcode scanning, nutrition lookup, pantry management, recipe discovery, and meal planning often live in separate tools. FoodFinder was an attempt to turn those disconnected steps into one coherent desktop workflow.
A product journey built around what is actually in front of you.
Scan
Barcode scanning through OpenCV / JavaCV so the product starts from a real object, not a manual form.
Analyse
Nutrition information and pantry data pulled through a local data model designed for repeated use.
Discover
AI-assisted recipe recommendations tied to what has actually been scanned and stored.
Plan
Meal and grocery thinking built into the flow so the product becomes more than a lookup tool.
Explore
Recipe Roulette and dietary filtering to keep discovery useful instead of static.
Layered on purpose.
Client
Java Swing desktop interface with scanner flow, pantry views, recipe exploration, and local-first interaction design.
Application
State management, scan handling, pantry logic, nutrition flow, recommendation orchestration, and product rules.
Services / AI
Computer vision and barcode decoding with optional AI-assisted recipe recommendation logic layered on top.
Database / Storage
SQLite-backed pantry, nutrition records, recipe rows, and offline-first persistence.
External systems
Product data inputs, recipe sources, and AI integrations used carefully rather than as the primary dependency.
Java Swing, SQLite, OpenCV, JavaCV, barcode decoding, structured pantry data, and AI-assisted recipe workflows.
The goal was not just information retrieval. It was a usable flow from scanning to decision-making, with enough clarity that a desktop product could feel helpful instead of dense.
Computer vision is used where it adds real value: getting the product into the workflow quickly by turning a physical barcode into structured application data.
AI sits on top of the structured pantry and nutrition data. It is used to improve discovery, not to replace the product logic underneath it.
SQLite keeps the product local, fast, and dependable. That choice made pantry state, nutrition data, and recipe flows easier to manage without inventing backend complexity.
The product is designed to stay useful even when external services are unavailable. That keeps the core desktop experience more resilient and more honest.
- Balancing camera-driven input with a desktop workflow that still feels reliable.
- Keeping AI as an enhancement rather than letting it become the only product story.
- Designing data flow and UI structure so the app stays understandable as features grow.
- FoodFinder became the strongest example of product thinking across desktop, data, AI, and computer vision.
- It reinforced that the best technical demos are the ones tied to a clear user flow.
- It also proved how much stronger a product feels when the architecture is shaped early instead of patched late.