back
user profile
brandonb
9,344karma·1,335submissions·January 17, 2011
about
Data for good.
Co-Founder at Empirical Health (https://empirical.health). Don't die of heart disease.
Before: Co-Founder @ Cardiogram (ML for heart health)
CTO at Sift Science (YC S11, machine learning to fight fraud)
Data Science @ UCSF Cardiology
HealthCare.gov rescue team
Google (Android speech recognition, search ads ML)
brandonb.cc
recent activity (1,335 total)
comment
Andrew Ng has a nice quote: “Instead of doing AI, we ended up spending our lives doing curve fitting.” Ten years ago you'd be ashamed to call anything "AI," and say machine learning if …
comment
Both this and Jesse's articles are great. Thanks for posting!
comment
Can you explain how? If I'm understanding the paper right, the timeseries encoding is a Conv1D and the cross-attention layer is constrained to output the token space of a pre-trained LLM. My naiv…
comment
This is very cool! From the paper, this technique seems to work well for question answering in time-series. In medical AI, IMO, the most exciting work is detecting disease signals too subtle for human…
comment
Glad you're still with us! (And that you're seeing a cardiologist soon.)
comment
The closest thing I've seen is the antioxidant index from Samsung's latest Galaxy Watch: https://www.mobihealthnews.com/news/samsung-debuts-galaxy-wa... It specifically…
comment
FWIW, the linked article shows a specific biomarker that would change both medication and lifestyle, i.e., something you’d act on. The panel from the company above (I’m a co-founder) includes a review…
comment
The most recent evidence is that if you’re measuring ApoB and Lp(a), you’re capturing the same information as fractionation (with the bonus that the former two are slightly faster and cheaper at most …
comment
My favorite definition of heart disease is "when circulating fats in the blood (lipids) are pushed by a driving force (blood pressure) into a vessel wall that is vulnerable (endothelial dysfuncti…
comment
Hi! Happy to help -- can you send me your debugging code (under Settings) if you haven't already? I'm bmb@empirical.health.
comment
Their comprehensive men’s panel is $200, which seems like the most comparable package. (The list of biomarkers above is a subset of the ones in our panel—obviously if you remove biomarkers, you’ll end…
comment
I updated the above post to say (OP here). I thought it was fairly clear from my phrasing of "we offer the tests as cash pay", but never hurts to be even more explicit. I checked privatemdla…
comment
Fair point!
comment
(OP here) LDL is still a good biomarker, but ApoB is a better biomarker for the same undelrying risk factor -- each atherogenic particle (LDL, VLDL, IDL) has exactly one ApoB molecule. The reason we o…
comment
Do you have a link to an article that covers the history? In the time that I've been following this topic (about 10 years), hs-CRP has been the go-to biomarker of inflammation. It'd be inter…
comment
The post links to the JACC article (which is an American College of Cardiology consensus statement).
comment
Note that the $190 panel includes not just hs-CRP ($59), but also the other major heart health biomarkers: ApoB ($69), Lp(a) ($49), A1c ($39), lipid panel ($59), eGFR ($99), other biomarkers, and a vi…
comment
I certainly get the skepticism. You might find the JACC article (which is an American College of Cardiology consensus statement) interesting, since the ACC is a neutral party here (and reviewed all th…
comment
Chronic inflammation (a general systemic thing) is the thing we're mostly concerned about for heart health. The hs-CRP metric itself will pick up both acute inflammation (if you get sick) and chr…
comment
Have you measured your Lp(a)? It's the strongest hereditary risk factor for heart disease. (Each Lp(a) particle is essentially a "normal" cholesterol particle with an extra protein that…
comment
Recommended in addition to cholesterol (or better yet, ApoB). And recommended for everyone, not just those taking statins.
comment
It's more the former -- we've gotten so good at detecting high cholesterol and reducing it, that the majority of residual risk is now in the other factors. (There are some people who dispute…
comment
Interesting! That makes a lot of sense to me -- and I could see how Korey could get better over time at interacting with other agents (e.g., with Devin to actually fix a bug).
comment
I suspect Lp(a) is next, since there are now drugs in clinical trials that directly lower it.