for those who don't know me, I tried to create my own company in the summer
of 2024. odu was a
platform to automate medical billing for physicians in ontario with a
rag
system.
funny story, I got this idea from my grandma's family doctor, call him john.
john and I were having small chats at my grandma's appointment, at the time,
gpt came out just a couple months ago. he asked questions about me,
naturally I
would ask back. john talked about a lot of admin stuff he needs to do and it
wasted a lot of his time, and jokingly said I should make one of those ai
things for him to do these tasks, and he'd pay me some money.
the way doctors in ontario gets paid is very weird, this is the case for
some other provinces and states as well. doctors would generally diagnose
the patient with something, and perform some sort of procedure if needed.
let's call diagnosis x and procedure y.
doctors gets paid through the
government
, and to get paid, the government wants doctors to tell them the diagnosis
and procedures for their patients. but the government don't want those in
words, they want it in codes. so perhaps a patient has flu,
their code could be something like 923. without submitting these codes,
doctors don't get paid. problem is, the definitions for these codes are so
long and they have different codes for different diciplines, e.g) radiology,
family doctor.
here
is a ~1000 page document that outlines all of the codes with their
corresponding diagnosis and procedure codes. it's clear how this becomes a
pain in the arse. most solutions out there were
companies hiring billing specialists
that charge a small percentage
of the doctor's appointment in exchange for finding out the codes and
submitting it to the government for them. this is actually a lot of money
when accumulated over time. so I wanted to build a cheaper and more reliable
alternative
I finished the mvp pretty quickly; and I sent emails to doctors at local
clinics and even walked in to some of them to find decision makers to talk
to. luckily for me, my first ever cold email got my first ever user. when I
looked at the supabase dashboard,
seeing a new signup made me feel so happy. I kept working on my product, and
every week or so I would check in to see what features my user really cares
about. then it came: doctors want a platform where they can send the codes
directly to the
government(the default government platform is super ass and hard to
navigate). after doing some researching, I need to make sure my software
passes some sort of conformance testing, which makes sense because the data
needs to be secure.
there's some guy on github that actually wrote a
code
that can pass this, so I was going just use that and host it on some vps. at
the time I thought to myself, holy shit I'm gonna create a profitable
company. little did I know, when applying for the conformance testing,
there is a point on the
form
that says everytime you make a change to your software, you need to REAPPLY
for conformance testing AGAIN. for a clueless guy that was following every
yc advice, this was killing me. one of the startup "rules" was that
you need to move fast. with this policy in place, I would not be able to do
any of that. and a lot of my joy comes from me shipping/improving features.
I thought to myself for a while, and decided to wrap things up. I was
already at a bad place doing business
in a domain that I wasn't familiar with, and with the policy I can't see
myself keep going.
on a side note, I had talks with vcs and startup accelertors (not those
super beefy ones you'd hear about on hackernews), and they were interested
in what I was building. though it didn't matter anymore, I wrapped this shit
up.