Work

Problem, build, result.

Client projects first, then the production systems and research behind them. Employer work is described without internal names or screenshots.

Client projects

Built for clients through QANTARA

Retail · WhatsApp AI assistant and operator console

Problem: An automation workflow was buckling under message volume across WhatsApp, Instagram and Facebook.
What I built: A bounded Gemini assistant with verified webhook ingestion and idempotent writes, plus a Spanish operator console with human handoff, follow-ups and a compliance log.
Stack: Node.js · Fastify · PostgreSQL · Redis · BullMQ · Gemini · Next.js

250+

conversations a day in one pipeline

Financial planning firm · Querétaro

Problem: A financial planning firm needed a site that explains its insurance plans and turns visits into booked consultations.
What I built: Plan pages with clear conversion paths and a Calendly booking flow for free consultations.
Stack: Next.js · Vercel · Calendly

Live

site and booking funnel in production

In-house at a Fortune 500 manufacturer · academic research · ventures

In-house engineer · Fortune 500 manufacturer

Problem: Engineers spent up to two days writing each technical document, and an earlier tool rollout had failed.
What I built: Sole engineer on a C# modular monolith that grounds generation in tag-based parsing of the source before multi-agent enhancement on AWS Bedrock. Serves ~50 engineers.
Stack: C# · AWS Bedrock · Multi-agent · Modular monolith

2 days → 2 hours

per document

In-house engineer · Fortune 500 manufacturer

Problem: Error emails from 15+ legacy servers across four regions were scattered, so failures were caught late.
What I built: A real-time pipeline that normalizes those errors into one stream and a dashboard operations teams can act on.
Stack: Python · SQL Server · MS Graph API · Power BI

−60%

unplanned downtime

Published research · IEEE CHILECON 2025

Problem: A dairy farm had no live view of milking-queue occupancy or its peak-demand patterns.
What I built: A YOLOv9 computer-vision system on a Raspberry Pi 4 monitoring the queue, with 95% accuracy across 720+ daily captures at under 3 seconds latency.
Stack: Python · PyTorch · YOLOv9 · Docker · Raspberry Pi 4

IEEE

CHILECON 2025 publication

Co-founder & backend engineer · pre-launch SaaS

Problem: Gated communities in Mexico manage finances, access and operations across spreadsheets and chat groups.
What I built: The multi-tenant backend: per-tenant isolation, RBAC and audit trails, CFDI-compliant payments (SPEI, card, OXXO), and a 7-channel notification engine.
Stack: TypeScript · PostgreSQL · Go · asynq · Redis · Stripe · Conekta

Co-founder

backend owner since 2025

Accepting new projects

Twenty minutes is enough to tell whether a Reliability Review or a build makes sense.

Reply within 48h · English or Español