Why I built Feed America

I started Feed America in 2021 because the food-assistance directory most people see — when they're hungry, when their EBT card is empty before payday, when they're afraid to call a stranger asking for help — was a maze of stale listings and broken phone numbers.

The data already existed

The data existed. USDA's Food and Nutrition Service publishes the SNAP retailer database. HRSA publishes the Health Center Locator with all 18,000+ FQHC sites. State WIC agencies publish their clinic registries. Plentiful, ampleharvest, and Salvation Army aggregate community pantries. NCES publishes school meal site data.

But each source lived in its own silo, with its own freshness rules, its own login wall, its own update cadence, its own data format. A family in food crisis trying to find a pantry tonight had to know which silo to look in, which probably wouldn't have hours, which probably wouldn't be in their language.

Identity by EIN

Giving platforms, donor-advised funds and the IRS identify a charity by its EIN, so we made Feed America’s EIN (92-1761881) part of every donor surface: Wikidata entity records, the EIN printed beside the legal name and mailing address, and the EIN on every gift instrument.

What "infrastructure-first nonprofit" means

I designed Feed America without staff payroll. We run on volunteer + AI infrastructure. Donor dollars go to platform infrastructure (Cloudflare Workers + D1 database serving 275,307 directory listings and 255,820 USDA-authorized SNAP retailers), data verification, community partnerships, and open-data publishing. Feed America files the IRS Form 990-N; see /transparency for filings and how funds are used.

This isn't ideology. It's pragmatism. A traditional nonprofit needs $200K-500K in annual fundraising just to cover salaries before any program work happens. By starting with infrastructure-as-code and AI-assisted data work, we shipped the same impact with a fraction of the overhead. The trade-off is real: I'm one person operating in volunteer capacity. We can't take grant requests that require dedicated grant-management staff. But we can do the directory work at scale.

Five years in, what's changed

In 2021, the directory had a few hundred entries I'd manually copied from public sources. Today it has 275,307 directory listings and 255,820 USDA-authorized SNAP retailers across all 50 states, DC, Puerto Rico, and U.S. territories. We operate public food-help search and publish interfaces for referral services, configured AI clients and researchers to evaluate. This does not establish third-party adoption. We have full Spanish-language parity. We earned the 2026 Candid Platinum Seal of Transparency — Candid's highest tier — in five years, faster than many comparable nonprofits.

What's still hard: operator-claim rate. Few listings have a verified operator claim so far. Each claim adds a one-tap "open today / closed today / low stock" status update for searchers. Closing that gap — getting more operators to take 60 seconds at feedam.org/pantry — is the highest-impact data-quality lever we have.

What I want next

Hunger is solvable. The infrastructure to find help just has to exist where people already are: on Google, on Wikipedia, on AI assistants, in 211 networks, on hospital discharge planning forms, in school counselor toolkits. We're building toward that.

If you operate a pantry, please claim your listing. If you have an integration use case, our APIs are open. If you can donate, list our EIN (92-1761881) on your gift to make sure it reaches us. If you'd rather volunteer than donate, we have five paths at /volunteer.

— Sharika Parkes
Founder, Feed America
Houston, TX