Amari AI

Amari AI — A large corpus of AI-drafted customs entries with their provenance trail, built across multiple leading brokerage teams: automated extraction from commercial invoices, packing lists, bills of lading and arrival notices, mapped through to ABI-ready draft entries with HTS classification and duty computed, each field traceable to its source document in an audit-ready trail - so the corpus pairs real source documents with the accepted structured entry. Layered on top is a continuously updated regulatory rule-mapping engine that tracks current tariff regimes including Section 122, IEEPA…

Amari AI offers (Alternative, Supply Chain, Fundamental) — Over 1 million customs entries automated, stated on the page - the strongest published volume anchor in this cohort. Each entry carries extracted document fields plus its structured output, so the effective training-record count is substantially higher than the entry count, plausibly tens of millions of document-to-field pairs. Brokerage tenant count and per-entry document volume are not published and should be confirmed.

Training and evaluating document-AI and OCR models on genuine trade documents paired with accepted structured entries - the strongest fit here, because at over a million entries the corpus spans enough document variety and formatting noise to be a real training distribution rather than a curated sample; HTS classification model training validated against broker-accepted entries; entry-error and pre-validation failure modelling to predict which filings attract examination or penalty; a dated regulatory-encoding series showing how successive tariff measures flowed through to filed entries, useful for measuring policy transmission into compliance behaviour; and importer-level and commodity-level flow signals aggregated across the brokerage book

The data is with 2 years of history.

Coverage spans US; Application Software, Air Freight & Logistics; alternative, supply_chain, fundamental; equities.