SiftyML

Sifty ML Ltd, operating as Sifty, develops machine-learning software for customs clearance of merchandise in international trade. Its product ingests a broker's manifest document, analyses more than 30 declared shipment variables — including HS code, descriptions, origin and destination countries, consignee and shipper details, weight, dimensions, value and restriction status — and automatically flags parcels whose descriptions do not conform to local regulations or whose data contains inconsistencies that could cause processing delays. Users edit the highlighted problem shipments and download the updated document, a three-step workflow the company positions as converting clerks into supervisors of data rather than manual inspectors [0][1].

The company frames its approach as machine learning used as a support system: computers take on low-cognitive, repetitive tasks while customs teams handle work requiring judgement and experience, allowing an organisation to process more volume with the same headcount [2]. Continuous user feedback is fed back into the system to keep classifications and recommendations aligned with current regulations [0][1]. A third-party company profile describes the product as sorting through thousands of parcels in seconds using patterns across declared data drawn from a base of more than 10 million transactions verified by authorities and customers [4].