My first AI API — and the app I built around it
I'm Kiril Sierykov, an 18-year-old developer. stackCount started as a portfolio project to prove I could ship a real full-stack app — and connecting my first AI API was the hardest part. Getting the OpenAI GPT integration working took real trial and error: API keys, prompt design, structured JSON output, caching responses, and handling failures gracefully. I stuck with it, figured it out, and now the Analyze feature sends 90 days of transaction data to GPT and returns actionable financial insights. That experience is exactly why this project belongs on my resume.
OpenAI GPT API integration
Production integration with gpt-4o-mini — structured JSON insights, response caching, and error fallbacks.
Full-stack architecture
Next.js 15 App Router, PostgreSQL via Prisma, NextAuth (Google OAuth + credentials), Stripe donations.
Shipped end-to-end
From database schema to deployed Vercel production — auth, payments, AI analysis, and CSV export all working.
