Case Study
Setwise
AI-powered football card marketplace
7,000+
Cards Catalogued
£2M+
Sales Tracked
13
AI Tools
I collect football cards. Yes, really. The market data was everywhere - eBay, forums, price guides. Fragmented. Impossible to track. I wanted something that didn't exist: real-time market intelligence without hiring analysts.
Built an autonomous market intelligence system. AI identifies popular cards, aggregates comprehensive data (trends, investment analysis, player profiles), and caches results for instant API responses. 13 custom AI tools working together. What would take analysts hours runs in the background daily.
Autonomous Market Intelligence
AI agents that run daily, identifying trending cards, aggregating comprehensive market data, and caching results for instant API responses. What would take analysts hours happens automatically in the background.
13 Custom AI Tools
Specialized tools working together: trend analysis, investment scoring, player profile generation, price prediction, rarity assessment, and more. Orchestrated through Mastra for reliable multi-agent coordination.
Real-Time Price Tracking
Aggregates sales data from eBay and other sources. £2M+ in tracked transactions. Historical price charts, market trends, and investment indicators for every card in the catalogue.
Sub-50ms Response Times
Intelligent caching layer ensures instant responses. Connection pooling, circuit breakers, and health monitoring provide production-grade reliability. 70% cost reduction vs. manual analysis.
Comprehensive Card Catalogue
7,000+ football cards catalogued with player profiles, card variants, rarity scores, and market valuations. Searchable, filterable, and continuously updated by AI agents.
7,000+ cards catalogued. £2M+ in tracked sales. Sub-50ms response times. 70% cost reduction vs. manual analysis. I don't just tell people how to build. I build for myself too. Same methodology. Same standards. Same speed.
Old me would have hired a team. Spent 6 months. £50K+.
New me: I'll just build it myself. 30 days. Solo.
The hardest part was the AI orchestration - getting 13 different tools to work together reliably. Connection pooling, circuit breakers, health monitoring. AI agent infrastructure needs the same reliability patterns as traditional microservices.
That's what I learned. And that's what I apply to client builds now.
Core Platform
Card catalogue, search, filtering, PostgreSQL database architecture
AI Integration
Mastra framework, 13 custom AI tools, market intelligence agents, trend analysis
Infrastructure
Connection pooling, circuit breakers, health monitoring, caching layer
Launch
Performance optimization, sub-50ms responses, production deployment
Building in Public
Follow along as I build tools, ship products, and share what actually works.
No spam. Unsubscribe anytime.