Shami Insurance Brokers

Shami, a Jordanian insurance brokerage, was buried in manual work. Policies arrived as paper, PDFs and scanned images with no structured record. And because Jordan has no online insurance aggregator, every customer quote meant emailing more than ten insurers the same car and driver details, then waiting on ten separate replies. Slow, repetitive and impossible to scale.

Two builds. First, an OCR pipeline that reads any incoming policy whatever its format, identifies which insurer issued it, extracts every field, customer name, vehicle make and model, sum insured, cover type, dates and more, and writes it to a shared online spreadsheet the whole team can work from. From there it can automatically send those customer and vehicle details out to the different insurers, then read the replies as they land in the inbox, compare them, and tell the team which insurer came back with the best price, along with any other notes the insurers included. Second, I reverse-engineered the insurers' pricing from their historical policies, mapping how car model and type, hybrid or electric, driver age, area and cover level drive each insurer's premium, and turned it into an in-house quoting tool, effectively the aggregator the market didn't have.