I used AI to build a free off-market company search using PPP. 822,220 companies, estimated financials, no paywall. Tear it apart.
redactedA post from Emile Petrone about using AI in a search stuck with me, especially the point that roughly half of deals never hit a listing site. The off-market half is where proprietary deal flow lives, but building a target list from scratch means stitching together data most of us buy piecemeal or scrape badly.
So I built the tool I wanted and I am opening it up free to this community: redacted
This should help anyone who is not Github tech savvy or does not want to build their own tool.
What it does today
- Search 822,220 US companies from the SBA PPP public loan data (150K and up file, final release)
- Filter by NAICS, state, city radius, employee count, and estimated SDE
- Every company shows estimated revenue and estimated SDE as ranges with a confidence label. The estimates come from payroll data and industry benchmarks, and the app tells you exactly which numbers are observed and which are modeled. Nothing is presented as fact that is not one
- Save searches as universes, build lists, watch companies, compare targets side by side, export up to 10,000 rows to CSV
- Market view: employee, geography, industry, and SDE distributions for any universe you save
The AI part, with a caveat I care about
You can attach your own Gemini or Claude API key and the app will scan the major listing sites for active listings that resemble a company profile, and let you chat about any company or market. Your key stays in your browser and calls go straight to the provider, never through my servers. A scan costs about 4 to 7 cents on your own key. I wrote up exactly how key handling works on the security page because you should not have to trust a stranger on a forum.
What it is not
The PPP data is a snapshot. Payroll figures date to 2020 and 2021, and the SBA published its final refresh in lateredactedIt is a discovery layer, not diligence. SBA 7(a) and 504 quarterly data and Form 5500 are next on the roadmap, which will add businesses that are borrowing and operating right now. I am working on adding additional data sources including local search.
The ask
It is free, there is no upsell behind it, and I want it beaten up by people who actually run searches. What data sources would change your sourcing? What filters are missing? Where do the estimates feel wrong for industries you know well? There is a feedback tab on every page, or comment here and I will respond.