Skip to content
tren
№ 003
live

Export AI

Export intelligence for manufacturers — market discovery, an "Export Fit Score", HS code analysis and multilingual outreach.

Ownership
Company · team · AI PORT
Role
Technical lead — built the frontend architecture from scratch
Team
4 developers
Period
Jan 2025
Open link
Live ↗
https://aiport.tr№ 003
Export AI — web

Context

For a manufacturer, exporting is the question "which country, which buyer, what first sentence in which language". Export AI reduces the three to one flow: market discovery on a world map, a company-specific Export Fit Score, HS-code-based demand analysis, and multilingual automated outreach to the resulting buyer list.

Constraints

  • Trade data is multi-source and noisy (providers such as Volza and ImportYeti); it cannot be scored before normalisation.
  • The score must be explainable: "why Germany, why now".
  • Outreach needs reputation management; bulk sending must not land in spam.

Architecture

From data to first contact: source → score → map/list → email.Trade dataHS code analysisExport Fit ScoreWorld mapBuyer listMultilingual email
From data to first contact: source → score → map/list → email.

The frontend is a static export on Next.js 14 App Router; module state lives in Zustand. The backend is shared with FLY AI: Java 17 / Spring Boot 3 and MongoDB.

TODO: To be extracted from the source: data-provider integrations, the scoring model, the email infrastructure (provider, queue, reputation).

§ Decision log

DecisionAlternativeWhy
01A company-specific Export Fit Score — one number, with explainable componentsPresenting raw trade data as tablesManufacturers do not employ data analysts; the decision must read at a glance, yet it must be possible to see why the score went up.
02Make the HS code the entry pointFree-text product searchThe customs tariff code identifies a product independently of country; market data and buyer lists join cleanly on that key.
03Multilingual email generation and sending inside the platformExport the list and hand it to an email toolFirst contact is the link that breaks most often in exporting; keeping generation, translation and sending in one place makes conversion measurable.
04Next.js 14 App Router + TypeScript + Zustand, shipped as a static exportA server-bound SSR applicationThe three visualisation-heavy modules (map, score, filtered lists) live on the client; static export simplified deployment, and Zustand kept state within module boundaries.

§ Outcome

−50%
Market discovery time (aiport.tr)
20–30%
Export volume growth (aiport.tr)
3
Interactive visualisation modules

§ Retrospective

What I would do differently today

TODO: What would you do differently today?