// The Marapone Process — Logistics
We don't sell software you figure out yourself. We design, build, and install a private AI system around your specific trade documents and carrier network — on your hardware, in your environment, with a full handover when it's done.
Submit one real document from your operation — a carrier invoice, bill of lading, customs entry package, or freight quote. We analyze it with our AI, review the output personally, and return a written report plus short video walkthrough within 24 hours.
No pitch. No obligation. Just a demonstration on your actual data so you can make an informed decision before committing to anything.
After the assessment, if we're a fit, we run a focused discovery session — usually 30–60 minutes — where we map your document types, volume, trade lanes, and the specific problems you need solved. No templates. No assumptions.
From that session we produce a written scope and fixed price. You know the full cost before signing anything. No surprises mid-build.
Document volume
How many invoices, shipments, and customs entries you process per month drives the architecture — we build for your actual load, not a generic assumption.
Trade lanes & commodities
Your specific lanes, HS codes, and carrier relationships determine how the AI is fine-tuned. A tech importer and a bulk commodities broker get fundamentally different builds.
Infrastructure & deployment
We establish where the system lives — your server, a local workstation, or an air-gapped environment — and confirm hardware requirements before scoping.
We build the AI system around your actual documents — not generic training data. The model is fine-tuned on your invoice formats, your carrier confirmations, your customs document types. It learns the patterns that are specific to your operation.
All fine-tuning happens in our secure build environment. Your documents are deleted from our systems after the build is complete and verified. Nothing is retained.
We install the system on your hardware, run it against a live sample batch to verify accuracy, and hand over everything — source code, model weights, documentation, and a team onboarding session.
After handover, the system is yours outright. You can modify it, extend it, or have another developer work on it. There are no usage restrictions, no subscription renewals, and no dependency on Marapone's servers continuing to exist.
"We don't take on many projects at once. When we're building your logistics AI, you have our direct attention — not a shared project manager. Typical build time from scoping to handover is under two weeks."