Services
Five things I build. Every one links to a real system I've shipped — proof, not promises.
[+] full-stack products
Whole software products, end to end
The full loop from one person: web app, mobile, or native macOS; backend, database, auth, billing, onboarding, distribution. I've shipped and sold my own product (napcat.app — native Swift, Homebrew, paying customers), which is the strongest guarantee I can offer that yours will actually ship.
[+] ai voice systems
Voice agents that answer real phone calls
An AI receptionist that answers 24/7, books appointments against real availability, detects emergencies and hands off to a human, and chases the leads that didn't convert. I've built this at production scale for home-service businesses — ~83k lines of it, including the workflow editor that lets owners change the behavior themselves.
[+] ai products & agents
AI features and agents that survive production
LLM-powered products and agents designed from the incident you fear backwards: scoped credentials they can't leak, idempotent actions they can't duplicate, destructive steps gated behind confirms, and a human approval stage with no code path around it. If your product should be callable by AI agents (MCP), I build that too.
[+] workflow automation
LLM pipelines for the drafting bottleneck
Somewhere in your operation someone spends hours producing first drafts of something formulaic — replies, outreach, quotes, summaries, triage notes. I build the pipeline that drafts it with your data as grounding, routes it through human approval, and never acts alone.
[+] internal tools
Internal tools & ops automation
The dashboard your ops team is missing, the API-to-API sync someone runs by hand, the approval queue living in a spreadsheet. Built in your stack, wired to your data, boring where it should be boring: real queues, idempotent jobs, audit trails.
how it works
faq
+What do you actually specialize in?
Shipping whole products: full-stack web apps, mobile apps, native macOS tools, and AI systems — voice agents, LLM pipelines, automations — plus the plumbing (queues, auth, billing, dashboards) that makes them real. TypeScript/Node ecosystems mostly, native Swift when the product calls for it.
+Which models and stacks do you work with?
Model-agnostic by design — I've shipped systems routing across Anthropic, OpenAI, and Google models, and pipelines with offline mock providers for deterministic testing. Stack-wise: Next.js/React, Node/Bun, Postgres, Supabase, and whatever you already run — I fit your stack rather than importing mine.
+How do engagements work?
Project-based with a fixed scope from the one-page spec, or a weekly arrangement for ongoing work. Tell me the problem and I'll propose the smallest engagement that solves it.
+Can you work with our existing codebase?
Yes — most client work is inside existing systems. My two longest client engagements each run hundreds of commits deep in codebases I didn't start.
+Why hire a freelancer for this instead of a vendor?
Vendors rent you their product's shape. I build the system in your shape and hand you the keys: your infra, your data, your team able to change it. When the work is done, you own an asset, not a subscription.
start
The first step costs nothing: describe the problem, and I'll reply with the four scoping questions and an honest read on whether AI is even the right tool for it.