About Me
I'm Brian. I am writing to you on Oct 18, 2024.
Hobbies
Nowadays,
-
I'm in the middle of my first Fantasy Football season, and it's going okay. I'm like
4-2 W-L right now. The real fun was I
wrote an ML routine to rank players for the draft. It worked! Much better than I
could have done on my own recognizance, and more pro-social than just clicking
autodraft.
-
I get a lot of reading in. You can find my reading list
here. I try to get a 50-50 split of fiction and non-fiction books.
-
I'm in a crew that does a weekly trivia night. We're good! We pretty reliably bring
home the 25-bar-dollars prize.
-
I like rock climbing, though it's hard to get out to it too often. It helps that my
early-grade-school kid is old enough to go, especially if we're meeting another family
at the gym.
-
And I'm very slowly spinning up some simple, ad-free, in-browser games my kids
can play on the iPad, like this Lights Out
game.
Formerly,
-
Skiing is a lot of fun, but it's hard driving up to Tahoe from Santa Clara county.
-
I did some beginner-level glass blowing. To show for it, I have a pair of little cups
and a large, wonky vase that do their jobs just fine. It's really fun, and I hope to
pick it back up again. Maybe even make something someone else would want in their home!
-
And I used to have a three-movies-a-week habit. Netflix was
mailing me discs as recently as
Oct 2020. Noir and crime worked especially well on me.
Find me on the internet:
Fun Stuff I've Made
- BlueSky Bots:
- Twitter Bots:
Read about how I made a bunch of bots for Twitter, including
@each_wordsquare,
@fakespearean, and
@ellipsabelle.
- My Wordle: The online word
game Wordle lets you share
a little green-yellow-grey tile pattern that reflects how you did on today's puzzle.
(You know the ones; they were all over Twitter; it's why the game was such a hit.)
I generated the average color of each tile over a history of 160-odd puzzle results.
It looks like
this. (Apr. 2022)
-
Wiggly Distributions and Nonparametrics: A blog post that exploring what
happens if you apply a nonparametric estimator to data drawn from discontinuous
distributions. (Jan. 2019)
-
WAVProject: You can take a
short snippet of audio, and repeatedly layer scaled-and-time-shifted copies of it. If you
do this enough times, with the right scale factors applied at the right time-shifts, you
can make the big pile of short snippets start to aurally resemble one much longer snippet.
(You're projecting the long .wav file onto the basis defined by the short
.wav file, is where the name comes from.) The effect this makes is pretty cool. You can
hear a demo here, where the
short snippet "BUILD" is used to recreate a longer snippet, "THE MACHINE."
(Dec. 2016)
-
Dice Roller: Gentle lil' JavaScript exercise
to roll "four eight-sided dice" or "three six-sided dice", etc., from back when I played
D&D. Mostly now use it to randomly pick albums to play from my Spotify library.
(May 2015)
-
UUID Baby Names:
A goofy utility for turning the output of
http://www.uuidgenerator.net/ into a human-pronounceable name. (May 2013)
-
City Picker:
Selects an American ZIP code at random, proportional to that ZIP's population,
and gives you its city name. Inspired by the wide and diverse array of cities
found in Josh Boruff's
SkyMall reviews. (May 2012)
-
Mandelbrot Oscillations:
A one-minute video demonstrating the recursively-calculated magnitudes of the iterations
of the Mandelbrot set's complex
quadratic polynomial, as a function of different values of c. (Jun. 2011)
-
BART Fares and the Triangle Inequality:
Finding routes on BART where it's actually more expensive to take a single, direct train
than to take the train part way, leave the system, re-enter, and finish the route on a
second train. (Jun. 2011)
-
Get Drunk But Neither Broke Nor
Fat: Considering calorie count and costs of alcoholic beverages. (Sep. 2010)
-
69 Love Songs, Illustrated: “Roses”: My contribution to
the joint effort to illustrate all the songs from the Magnetic Fields' triple album.
You can check out the video version
on Youtube. (June 2010)
-
RISK STATS: An almanac of the probable outcomes of battles in
the game Risk. (Feb. 2010)
-
The New York Times Columnist Comparator: A list of a few
distinguishing keywords from the regular columnists of the New York Times Op/Ed section.
(Updated daily from Oct. 2008 to Aug. 2009.)
-
The BoingBoing Filter: A list of a few distinguishing keywords from the
major contributers to BoingBoing. (Updated daily
from Oct. 2008 to Apr. 2009.)
Career & Education
(Here's my resume.)
I'm a quantitative software engineer (SWE) with Google, working on AI for managing your
Cloud Platform infrastructure. Can't say too much about it right now! (You can probably
guess what's coming, though.) One thing I can say is I am now a Golang developer. It's neat!
Previous Google teams: an analytical SWE helping optimize the various planet-scale storage
systems, an ML SWE on Google Fi's cell network selection & quality, and another analytical
SWE role working on the public web index.
Pre-Google, I was a data scientist at Upwork, working
on matching and ranking for their online labor marketplace. Before that, I was a data
scientist at Quantifind trying to mine social media comments
in a way that could predict the health of a product's marketing campaign.
I got my Ph.D. at the U.C. Berkeley EECS department
in 2012, working with Laurent El Ghaoui as
part of the StatNews research group. Before that, I got my B.S. in electrical engineering in 2005
from the University of Virginia, where I concentrated on statistical signal processing
with a computer science minor.
Publications and Conferences
-
Brian Gawalt. Convex Approaches to Text Summarization. Doctoral dissertation,
University of California, Berkeley. Dec. 2012.
[UCB EECS]
[eScholar]
-
Kurtis Heimerl, Kashif Ali, Joshua Blumenstock, Brian Gawalt, and Eric Brewer.
Expanding Rural Cellular Networks with Virtual Coverage. Proc. USENIX Conference
on Networked Systems Design and Implementation (NSDI'13), pp. 283-296. Lombard, IL, April
2013.
[Portal.ACM]
-
Kurtis Heimerl, Brian Gawalt, Kuang Chen, Tapan Parikh, and Bjoern Hartmann.
CommunitySourcing: Engaging Local Crowds to Perform Expert Work via Physical
Kiosks. Proc. SIGCHI Conference on Human Factors in Computing Systems (CHI '12).
Austin, TX, May 2012. [website]
[Portal.ACM]
-
Luke Miratrix, Jinzhu Jia, Brian Gawalt, Bin Yu, Laurent El Ghaoui,
Summarizing large-scale, multiple-document news data: sparse methods & human
validation, UC Berkeley Dept. of Statistics Technical Report #801, May 2011.
[abstract]
[pdf]
-
Saheli Datta, Brian Gawalt, Guan-Cheng Li, Luke Miratrix, Laurent El-Ghaoui,
Bin Yu, Abigail De Kosnick, Gaining Contextual Insights into Media from
Keywords Derived from Machine-Learning Based Analysis, Crossing Boundaries
2011, Berkeley, CA March 2011.
-
Brian Gawalt and Youwei Zhang and Laurent El Ghaoui, Sparse PCA for Text Corpus
Summarization and Exploration, NIPS 2010 Workshop on Low-RankMatrix Approximation,
Whistler, BC, Dec. 2010.
[pdf]
-
Brian Gawalt and Jinzhu Jia and Luke Miratrix and Laurent El Ghaoui and Bin Yu and Sophie
Clavier, Discovering Word Associations in News Media via Feature Selection and Sparse
Classification, Proc. ACM International Conference on Multimedia Information Retrieval
(MIR2010), Philadelphia, PA, Mar. 2010. [website]
[Portal.ACM]
Visit gawalt.com for more Gawalts.