Loupe

The problem

Finding one kind of business across several locations means bouncing between maps, websites and spreadsheets, then repeating the same manual checks for every row.

How I approached it

Separate discovery from enrichment. Search each selected geography independently, interleave and deduplicate results, then read public company websites for contact and company signals. Robots aware, and honest about what is inferred rather than sourced.

What I built

• React search interface streaming results over server sent events
• Map and list views with live result streaming
• Robots aware enrichment layer over public company websites
• CSV export and a proof led marketing site

Where it stands

Working internal prototype. It finds and records, and deliberately does not send outreach.

What it taught me

The interesting problem was not scraping. It was provenance. Every field needs to say where it came from, or the whole dataset becomes unusable the moment someone questions one row.