White-label
You sell it. I build it. Your brand keeps the credit. NDA-first — and I don't poach clients. Ever.
KEVIN GUIFARRO · PRODUCTION AI ENGINEERING · CST
Demos run once.Systems run every day.I only build the second kind.
Production-grade AI agents & integrations — n8n, Claude, TypeScript, RAG — shipped with evals, observability, and docs. Senior hands only.
8 YRS · FORTUNE 500 SENIOR ENGINEER5.0 ON UPWORK · VERIFIED
Three rules I don't break.
01
If it can't be measured, it didn't ship.
Every build carries an eval suite — accuracy, cost, latency. "It works" is a number, not a feeling.
02
If you can't see it running, it isn't.
Dashboards, logs, alerts. You — and your client — can watch it breathe.
03
The person who quotes is the person who codes.
No bench of juniors, no telephone game, no account manager.
Proof
Solo, on purpose
The engineer you talk to is the engineer who ships.
Timezone
Costa Rica, CST. Your 9am standup is my 9am standup.
Fortune 500
8 years keeping enterprise systems alive. Production doesn't care about your demo — I learned that the expensive way.
Stack
n8n · Claude & OpenAI APIs · TypeScript · RAG · Postgres · Docker
Delivery standard
Every build ships with: eval suite, observability dashboard, docs, handoff video.
Receipts.
Real systems, honestly labeled.
Flagship — Whis
Whis — my personal AI chief of staff, in production. n8n + Claude + Postgres. Google Calendar & Gmail via OAuth, Telegram interface, confirmation gates on every write action, stateful dedup so it never nags twice. It briefs me at 7:00 every morning — this is the system the headline is talking about.
Full engineering writeup — coming soon
Built to show range, labeled as what they are: demos.
Most of my white-label work ships under someone else's logo. That's the point. References available on request.
“Exceptional work — delivered fast, explained clearly, and it just works.”
Fixed price, fixed scope. Hours measure effort — you're buying outcomes.
Pilot Sprint
from $[X]
One agent or integration, working end-to-end in [2] weeks, evals from day one. The cheapest way to find out if we fit.
Book a 20-min fit call →Production Build
from $[X]
The full system: agents + integrations + RAG where it earns its place. Ships with evals, observability, docs, and a handoff your team can actually run.
Book a 20-min fit call →Keep It Alive
$[X]/mo
Monitoring, model updates, prompt-drift checks, cost tuning — and a human who answers when something breaks at 2am. Because it will.
Book a 20-min fit call →About
I'm Kevin. Before code, I fought — Muay Thai, kickboxing, BJJ. Competition teaches you two things: respect preparation, and don't flinch under pressure. Both turned out to be transferable.
Eight years as a senior engineer at a Fortune 500, building systems teams across the Americas depended on daily. Now I build AI systems to the same standard: if it can't run unattended on a Tuesday, it isn't done.
One-person operation, on purpose. The person who scopes is the person who quotes, builds, and answers. Based in Costa Rica, working US hours, fluent in English and Spanish.
FAQ
Why a solo engineer instead of another agency?
Because you're the agency. I'm the senior capacity you rent when the pipeline outgrows the team — without the overhead of hiring.
Do you work white-label / under NDA?
Default mode. Your brand, your client relationship, my code.
What does "production-grade" actually mean?
Evals that prove it works, observability that shows it running, docs that survive me getting hit by a bus. If a build lacks one of the three, it's a demo.
What DON'T you do?
Mobile apps, WordPress rescues, Microsoft Power Platform, and anything I'd have to learn on your dime. If it's outside my stack, I'll say so on the first call and point you to someone good.
Where are you, and when do you work?
Costa Rica — permanent CST (UTC-6). No DST here: my clock never jumps on your standups, year-round.
Blog
All posts →How I Built an AI Chatbot for a Restaurant in 48 Hours
A real breakdown of building a production-ready AI chatbot for a restaurant — the stack, the challenges, and what clients actually care about.
RAG vs. Fine-tuning: What Actually Makes Sense for Small Businesses
Most small businesses don't need fine-tuned models. Here's how to think about AI implementation without overengineering it.
I Built My Portfolio in 1 Hour Using Claude Code — Here's What I Learned
A developer's honest take on building a production site with AI-assisted coding. The good, the surprising, and what I'd do differently.
Send me the build your team doesn't have time for.
Or skip the calendar: kevin@guifarro.dev
WhatsApp: wa.me/506[NUMERO]