Build an MVP – AI-Powered Pokémon Card bargain Finder

Overview I'm looking for an experienced Python developer to build an MVP for an AI-powered web application that scans eBay UK for Pokémon card listings and identifies hidden bargains. The objective is not simply to compare prices. The objective is to identify listings that are likely to be undervalued or overlooked because they have poor titles, spelling mistakes, vague descriptions or require image recognition to identify the card correctly. This is an MVP and there is likely to be ongoing work after completion. Objective Create a system that continuously scans eBay UK and automatically: Finds Pokémon card listings Identifies the exact card being sold Retrieves current market prices Calculates whether the listing is undervalued Scores how likely the listing is to be a hidden bargain Displays the best opportunities in a simple dashboard Initially, the application should focus on single-card listings only. Technology Stack Preferred technologies: Python FastAPI (preferred) or Flask PostgreSQL GitHub Railway (deployment) Docker APIs & Services eBay Browse API Use for: Searching listings Reading titles Prices Images Listing URLs Item information Developer Docs: https://developer.ebay.com/ Pokémon TCG API Use for: Card database Sets Card numbers Images Rarity TCGplayer market prices Cardmarket pricing where available https://pokemontcg.io/ OpenAI API (Vision) Should only be used when required. For example: Poorly described listings Misspelled titles Generic titles Image recognition when the title cannot confidently identify the card The application should minimise AI usage to reduce running costs. Core Concept The application should NOT search only for specific Pokémon cards. Instead it should perform broad Pokémon searches and analyse every relevant listing. Examples: Pokémon card Pokémon TCG Pokémon holo Pokémon rare Pokémon collection Old Pokémon card Common misspellings Generic titles The purpose is to discover listings that ordinary keyword searches may miss. Functional Requirements 1. Broad Listing Discovery Continuously search eBay UK using broad search terms. Retrieve: Listing title Price Images Listing URL Item ID Initially: Buy It Now listings Single-card listings English language listings 2. Listing Analysis Analyse each listing. Determine: Is this likely a single card? Ignore obvious accessories Ignore sealed products Ignore bulk lots (for MVP) 3. Fuzzy Text Matching Implement fuzzy text matching using an appropriate library (e.g. RapidFuzz). The software should: Correct spelling mistakes Match vague titles Match incomplete titles Compare against the Pokémon TCG database Produce a confidence score Examples: Charzard ↓ Charizard Obsidan Flames ↓ Obsidian Flames Pikacu ↓ Pikachu 4. AI Image Recognition If fuzzy matching confidence is below a configurable threshold (e.g. 90%), analyse the listing image using OpenAI Vision. The AI should identify: Card name Set Card number Language Estimated condition (where possible) Confidence score 5. Market Pricing Retrieve market prices using the Pokémon TCG API. Compare: Current listing price Estimated market value Calculate: Discount (£) Discount (%) Potential profit 6. Hidden Bargain Detection This is the primary feature of the application. The software should identify listings that are likely to be overlooked by other buyers. Create a Hidden Opportunity Score using factors such as: Misspelled Pokémon names Misspelled set names Poor titles Generic descriptions Missing card number AI image identifies a better card than the title suggests Difference between asking price and market price Confidence of identification Example: Score 100 Extremely likely hidden bargain Score 90 Very strong opportunity Score 75 Worth reviewing Score below 75 Low priority Listings should be ranked using this score. 7. Dashboard Simple web dashboard displaying: Card image Card name eBay price Estimated market value Hidden Opportunity Score Potential profit Discount % Link to eBay listing Date/time found Sorting: Highest Opportunity Score Biggest discount Highest potential profit Newest listings 8. Background Scanner Continuously scan eBay at configurable intervals. Avoid duplicate processing. 9. Database Store: Listings processed Identified cards Market values Scan history Opportunity score Date found 10. Configuration Use environment variables for: eBay Client ID eBay Secret Pokémon TCG API Key OpenAI API Key No secrets should be hardcoded. 11. Deployment Application should be deployable to Railway. Please include: Dockerfile requirements.txt README Deployment instructions Nice-to-Have Features (Future Phases) Please apply only if you have experience with: Python REST APIs FastAPI or Flask PostgreSQL Docker GitHub OpenAI API eBay API (preferred) AI/LLM integrations Proposal Please include: Examples of similar projects. Estimated completion time. Any suggestions to improve the architecture or performance of the MVP. The goal is to build a robust MVP that can later be expanded into a full AI-powered Pokémon card discovery platform.