I build things that arescalable
Stop building at yesterday's speed. I combine full-stack engineering with AI-native workflows to ship high-end SaaS solutions and complex data tools faster than ever.
Featured Projects
Recent work showcasing full-stack development with AI integration and complex data systems.
AI-powered invoice management and SEPA payment platform for the European rental industry. Uses Mistral AI with intelligent model fallback: starts with a small model, validates quality, and escalates to larger models only when needed. IBANs are encrypted with AES-256-GCM, and serverless edge functions handle OCR and data extraction.
Complete rebuild of Ad Hoc Data's web platform across three refactor iterations: from legacy AngularJS to Angular 21 SSR, including a dual-database selection tool — KVK Pro (3.7M Dutch companies) and AI Leads (3.4M web-indexed companies across four countries) — with live record counts, 3,800+ technology filters, and dual export.
Multi-tenant CMS that lets non-technical business owners edit their custom websites. Built on a security-hardened AWS media pipeline: a fully private S3 bucket behind CloudFront Origin Access Control, presigned browser uploads, and least-privilege IAM — all defined as code with AWS CDK. A published npm SDK keeps every forked client site compatible.
Agentic AI content engine built for Ad Hoc Data's marketing: a daily GitHub Actions pipeline scrapes industry sources, scores inspiration with Gemini, matches it against CBS open-data statistics, and drafts a full content calendar — three social posts a week, a weekly blog, and a monthly newsletter.
My personal AI command center: an agent platform running 48+ Supabase Edge Functions with multi-provider AI (Gemini, Claude, Mistral, OpenAI), RAG knowledge bases on pgvector, a HubSpot OAuth2 app, and an analytics cockpit. Its public face is the Justin AI chat widget — running live on this site right now.
Building with High Velocity: AI-Native Workflow
I leverage Claude Code directly in my terminal to accelerate development without sacrificing code quality. By offloading the boilerplate to AI, I maintain my focus on high-level architecture and system design. This AI-native approach allows me to tackle complex logic and ship production-ready features at a pace that traditional workflows can't match.
Terminal-First Workflow
Claude Code integrated directly into my development environment for seamless AI assistance.
AI Agents & Automation
AI agents and pg_cron jobs handle the repetitive work automatically — from lead qualification in CrewVee CRM to nightly data pipelines running in the background.
Rapid Iteration
From concept to production-ready code in hours, not days. Complex features shipped faster.
// Secure AI invoice processing pipeline
async function processInvoice(encryptedFile: Buffer) {
// AES-256-GCM decryption
const decrypted = await crypto.decrypt(
encryptedFile,
await getFileKey(invoice.keyId)
);
console.log("✓ Decryption successful");
// Extract text with OCR
const ocrText = await extractText(decrypted);
// Mistral AI extraction with smart fallback
const result = await mistral.chat({
model: "mistral-small-latest",
messages: [{
role: "user",
content: `Extract invoice data:\n${ocrText}`
}]
});
// Validate & escalate if needed
if (!isValid(result)) {
return mistral.chat({
model: "mistral-large-latest",
// ... retry with larger model
});
}
// Re-encrypt sensitive data before storage
const encrypted = await crypto.encrypt(
result.bankAccount,
await generateKey()
);
console.log("✓ Encryption successful");
return { ...result, bankAccount: encrypted };
}