I build AI agents that actually run in production — not demos that break on day two.

I've built and operate an AI agent (Anthropic Claude API, 12 custom tools) that manages back-office operations for a pharmacy business, every day. I know what it takes to go from "cool prototype" to "the business depends on this": careful tool design, guardrails, error handling, and an eval suite that attacks the agent before every release.

What you get: — An AI agent connected to YOUR systems (database, APIs, email, WhatsApp, internal tools) — Custom tools designed for your specific workflows — Guardrails + adversarial testing, so it behaves when users don't — Secure server-side API key handling — Deployment + documentation so your team can operate it

Typical use cases: order management, customer support, internal operations, data processing, appointment handling.

Process: short discovery call → fixed scope proposal → build with progress updates every 2-3 days → deploy + handover.

Stack: Anthropic Claude API, Next.js, TypeScript, Supabase, Node.js.