The US has the most expensive healthcare system in the world. Finding a doctor in it should not feel like detective work.
A few years ago I moved to a new city and needed to find a primary care doctor. I spent hours on insurance websites that had not been updated since 2019, called numbers that were disconnected, and found doctors listed as "accepting patients" who had not taken new patients in three years.
The data existed. The government maintains a complete registry of every licensed US provider — millions of them, with addresses, specialties, and NPI numbers. It is public record, updated continuously, freely available to download. But nobody had built a useful interface around it.
The existing directories were either locked behind insurance systems, full of sponsored results that buried the best match under whoever paid more, or quietly selling your search behavior to pharmaceutical companies.
So I built okDoktor.
What I wanted to build
Something that felt like asking a knowledgeable friend — one who happened to know every doctor in the country. You ask in plain English and you get real answers, not a sponsored list dressed up as search results.
Something language-aware. Over 67 million Americans speak a language other than English at home. Finding a doctor who speaks your language is not a luxury — it is how you describe symptoms accurately and understand your diagnosis. That should be a first-class search, not a buried filter.
Something free for patients. Not "free with ads." Not "free until we sell." Searching for a doctor will always be free. That is non-negotiable.
How it stays honest
Search rankings are determined by AI relevance alone. No provider can pay to appear higher in results than they deserve. That is the core promise and it will not change.
Providers can pay to enrich their profile — adding a photo, bio, languages spoken, accepted insurance, and telehealth details. That is how okDoktor stays running. It makes their listing more useful. It never affects where they appear in your search.
Think of it like Google Maps. The map is free. Businesses can add photos and details. Nobody buys their way to the top of directions.
How it is built
okDoktor is a one-person project built with Rust on the backend — fast, memory-safe, designed to last. The search uses pgvector embeddings which means the AI understands what you are looking for, not just which words you typed. All of this runs efficiently without burning money on unnecessary infrastructure.
The NPI data is ingested in full, normalized, enriched, and indexed for semantic search. When a provider claims their profile they can add everything the federal record does not capture: languages, bio, insurance networks, telehealth availability.