PokeWealth

Upload Images
Collection of Cards
Portfolio over time

Inspiration

The Pokemon card market is worth $10B+, yet collectors have no easy way to track their portfolio value. PokeWealth changes that — just snap a photo and instantly see your card's worth, condition grade, and price trends.

What We Built

PokeWealth lets users upload a picture of any Pokémon card and instantly view its market value, condition, and historical price trends.

🧩 Frontend: Built with Next.js, TypeScript, and Tailwind CSS for a fast, responsive interface.
⚙️ Backend: Developed in Python using FastAPI for asynchronous REST endpoints and image processing.
🤖 AI / Vision: Leveraged Gemini 2.5 Flash to identify the card name, rarity, and condition through multimodal reasoning (image + text).
💰 Pricing Data: Queried live market data from the Pokémon TCG API and eBay API.
🗄️ Database: SQLite for card collections and price history

What We Learned

Multimodal AI integration: We learned how to feed structured image and metadata prompts to Gemini 2.5 Flash for more accurate card recognition and grading.

Async architecture: Implemented asynchronous workers to handle long-running image and API tasks without blocking the main FastAPI server.

Data visualization: Used charting libraries to display price history similar to a stock app, turning collectibles into actionable financial data.

Challenges We Faced

Image ambiguity: Similar-looking cards (especially alternate arts) often confused the model — we mitigated this using Gemini’s reasoning context and partial text OCR.

Pricing normalization: eBay and TCG APIs return inconsistent pricing data, requiring statistical smoothing (e.g., filtering outliers using a rolling median).

💡 Reflection

Building PokeWealth taught us that AI can make even niche hobbies smarter. We combined computer vision, cloud systems, and financial data into a product that makes trading cards feel like a real investment platform.