Landed Cost

SaaS that computes the true landed cost of every SKU imported into Mexico, reading the documents with AI.

Industry
International trade
Platforms
Web · API · AI
My role
Architecture, full-stack and mobile development
SKU · MXN

The challenge

Landed cost per SKU means reconciling the commercial invoice, packing list, freight bills, customs broker statement, import declaration and insurance. Doing it by hand in Excel is slow, and mistakes are expensive.

The solution

Users upload the shipment documents; AI classifies them and extracts the data, and a deterministic engine does all the math: incoterm, customs value, tax cascade, exchange rates and per-SKU allocation.

What I built

  • Multi-tenant API with Node.js, Fastify 5 and PostgreSQL 17, with system-admin and client roles.
  • Claude (Anthropic) extraction from document PDFs and photos, plus direct CFDI XML parsing.
  • Automatic exchange rates from Banco de México's API and manageable tariff rates.
  • React, Tailwind and Recharts dashboard, plus a bilingual Astro marketing site.

Technical details

  • AI reads and classifies; it never computes
  • Every number comes from tested pure functions
  • Multi-tenant: each client sees only its shipments
  • Swagger-documented API

Stack

  • Node.js
  • Fastify
  • PostgreSQL
  • Knex
  • Claude API
  • React
  • Tailwind CSS
  • Recharts
  • Astro

Need something similar?

Tell me about your operation and I'll tell you how I'd solve it.

Tell me about your project