I Gave Claude Code 9.5 Years of Health Data to Help Manage My Thyroid Disease

I have episodic Graves' disease, which has been difficult b/c its not chronic. Meds are up and down and often lag when the actual onset occurs I fed Claude 9.5 years of my Apple Watch and Whoop data, and tasked it to build an ML model (ended up with XGBoost after I tasked it to run every ML model, ran for over 1 hr) to detect these phases. It hit \~98% validation accuracy and now acts as a personal risk assessor, alerting me 3-4 weeks before symptoms even appear. Backtested it on my last episode, and it would've given me a heads-up in early August before labs confirmed it at the end of the month. I was pretty blown away by this, it even made some very novel approach shift decisions.  Turned it into a simple iOS app I can check whenever. I wrote this article given alot of interest I saw in emulating this along with the repo w/ claude code setup open sourced. Hope this helps