back
user profile
brandonb
9,303karma·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
"What is life" is one of my favorite books. It's packed full of fascinating explanations of how the most basic elements of life came to be. For example, did you know that sex evolved from cannibalism …
comment
I went to sign up as an employer, but stopped since I didn't want to fork over $50 before even seeing what quality of engineers you have. I'll bet you'd get a lot more employer interest if you let peo…
comment
One problem with finding a job is the paradox of choice. The more choices you have, the harder it is to make a final decision. That's especially true of startups, where there are literally thousands o…
comment
I may just be really tired and missing something obvious, but I see how to comment on other people's posts, but not how to post to the main roundtable. There's a message at the bottom of the page sayi…
comment
I originally tried to post this on the roundtable, but was discouraged by the requirement to "apply for membership". I'd love it if the roundtable just let anybody post, but allowed the moderator to f…
comment
Android 2.0 did integrate Facebook into the OS -- FB contacts became phone contacts, and when people called, you'd see their last status update. It was an awesome feature, but Google disabled it in An…
comment
The really unique thing about Siri is the flexible natural language processing. You can say "Meet with Doug for dinner today", and it'll parse out the person/date/subject, realize you didn't specify a…
comment
You've got it backwards. Modern speech recognizers have a vocabulary of a million words and multi-gigabyte models. It's generally much more accurate to do speech recognition in the cloud, since you ha…
comment
Moore's law suggests 4x performance increase over 3 years, and they doubled the number of machines, so you end up with roughly an 8x performance improvement from "just" hardware.
comment
Their benchmarks seem a little odd. They claim an 11x speed improvement, compared to using half as many computers three years ago. But you'd expect roughly an order of magnitude improvement anyway, ju…
comment
Ultimately, it's their money. If you were a VC, wouldn't you be re-assured to see the company being frugal?
comment
The black box can adapt to new circumstances. A lot of the algorithms used today in speech recognition were established on much smaller data sets (thousands of hours of speech); the tradeoffs made the…
comment
The really great thing about deep networks isn't that they're more accurate. It's that they're radically simpler. Current speech recognizers are basically layer upon layer of tricks discovered by rese…
comment
Sometimes you hear people say "we're like a startup inside of a big company." This is a great example of why it's hard to make that work. In a big company, you have one source of funding, and they can…
comment
There's a concrete reason you see very few Googlers working part-time with startups. Google has a conflict of interest policy that disallows employees from working on competing products; since Google …