I help companies ship AI that actually works in front of real customers — sales & support agents, automation, and the guardrails & evaluation that keep them honest. Not demos. Systems your business runs on.
Most AI projects die as demos. My focus is the hard part: making language models reliable, safe, and useful in production.
WhatsApp / chat agents for sales and support that quote from live data, answer accurately, and hand off to humans at the right moment.
Deterministic guardrails and evaluation loops so your model never invents prices, over-promises, or misleads a customer.
Process automation, ad-conversion & attribution pipelines, and integrations that connect the tools your business already uses.
End-to-end products: React/Next front-ends, serverless APIs, Postgres/Supabase — owned from database to shipped UI.
Real production systems a growing company runs on every day — not slideware.
LLM agents handling customer conversations daily, with anti-hallucination guardrails and a nightly automated audit.
Click-to-sale attribution for Google & Meta, plus a watchdog that caught a silent 12-day outage.
Inventory, quotes, orders, CRM and integrations — React, serverless APIs and Postgres, end to end.
Everything I know I learned by putting real AI in front of real customers, end to end.
I build with agentic tools daily, which lets one engineer move like a team — fast, without cutting corners.
My specialty is trust: guardrails and evals so the AI is safe to put in front of your customers.
Based in Bogotá (UTC−5). Work in English or Spanish, async-friendly across time zones.
Tell me what you're trying to build or automate. First call is free — we'll figure out if it's a fit.