When the US sanctions someone — a person, a company, a ship — the name lands in a giant government file, updated constantly and published as one 30-megabyte XML dump that takes minutes to download. Useless in raw form. The sanctions snapshot fixes that: a small server downloads the dump in the background, normalizes it into a tidy database, and serves a page where you type “Syria” and instantly see the counts, the programs, and the notable names, with charts.
The engineering is deliberately boring in the best way. The parser streams the XML instead of loading it whole, and it ignores XML namespaces so a government formatting change cannot break it. Refreshing is polite: if two people ask at once, the second one is told a refresh is already running instead of stampeding the download. The tests never touch the network — they run against a fixture file that happens to star a real sample entry, a Turkish coffee company under the Cuba program.
Students: this is data engineering in miniature — fetch, normalize, cache, refresh — and every backend job you will ever interview for is some version of it. Hiring managers: a fast cached API over an unusable government dump, with tests that do not flake, signals a builder who thinks about the next maintainer, not just the demo.
No comments yet