Outlier.Care is “levels.fyi for Healthcare”
Table of Contents:
Outlier.Care Demo
Three of my closest friends were finishing medical residency in the spring of 2023, and I was bored selling/marketing my to-be-announced startup everyday. I helped them evaluate their job offers, but there wasn’t much data to go off of except mildly useful averages that various medical groups print each year (AMGA, Medscape, Doximity, MGMA, et cetera 👀). I shared levels.fyi with them and asked if it existed for Healthcare. They unanimously said: No.
My background is in Product Management and I understand software concepts well, but I cannot program a line of code to save my life. I wanted to scratch my creative itch, so I thought to myself:
<aside> 💡 Could I use ChatGPT to build levels.fyi for Healthcare?
</aside>
Introducing: Outlier.Care
With Sahil Lavingia (@shl) lighting the spark:
https://twitter.com/shl/status/1645265642965315588
And Steph (@stephsmithio) & Shaan (@ShaanVP) providing extra motivation, I built a full featured physician compensation transparency app in less than 1 month. Here’s how it went:
Matching the ethos of levels.fyi, Outlier.Care will present user-submitted, granular compensation details without a paywall. But this necessitates physicians wanting to share their compensation, which itself necessitates physicians wanting to visit the site. The latter became the first problem I set out to solve.
What value can I provide physicians that they don’t get anywhere else? I realized that many job postings nowadays provide compensation ranges, but they’re tucked away in job descriptions and scattered across dozens of job boards. So that’s where I started.
Doximity is the “LinkedIn for Doctors” and has hundreds of job postings at any given time, many of which include compensation data (e.g. “$328-402/hour”). Instead of manually cataloging each job, I asked ChatGPT to automate it. In ~3 days, I had a fully functional web scraper that addressed the common pitfalls (pagination, sub-detail pages, IP blocks), and resulted in a CSV file containing 650+ job postings. Now, I know how to rotate IP proxies, user agents, and cookies to circumvent most anti-scraper defenses. See prompts below.