Why We Built VisaVigilAI

It started with ten addresses.

A partner here needed to know whether any of ten project sites qualified as a Targeted Employment Area. For an EB-5 investor, that answer decides whether the minimum investment is $800,000 or $1,050,000, so it isn’t a small question.

Nothing we could find would simply answer it. So we did it the slow way. One address at a time, pulling census tracts, checking unemployment against the national average, working out which neighboring tracts were allowed to count. Ten addresses took most of an afternoon.

Somewhere in the middle of that list, the obvious thought arrived: this is a script.

So we wrote one. What had taken an afternoon took seconds. That script is still running today — it’s the engine underneath the TEA checker on this site, which is also why the tool takes a whole list of addresses rather than one at a time.

Then we started wondering what else was sitting in plain sight, unusable.

Quite a lot, as it turned out.

The State Department publishes the Visa Bulletin every month and keeps the old ones online, so that data isn’t hidden. It’s just 130 separate web pages. Ask how fast EB-2 India has moved over the past decade and you’re opening all 130 and copying dates out of tables by hand. Other data really is gone. USCIS posts processing times that change without notice and keeps no history, so whatever that page said last August is unrecoverable. The Department of Labor’s PERM reports were scattered enough that we rebuilt years of them out of web archives.

That’s the thing about this data. You can research a question later. You can’t research a number nobody wrote down.

VisaVigilAI is what all of it turned into: 130 Visa Bulletins parsed back to 2015, every category and country, both charts. Daily snapshots of USCIS processing times across 43 forms and every office. PERM queue history back to 2020. Employer sponsorship records to 2009. It refreshes on its own, and every free tool on this site runs on it.

The TEA checker still does the thing it was built for. Ten addresses take seconds now. So does a spreadsheet with two hundred.