Julia Ferraioli (class of 2007)
Senior Developer Advocate
@juliaferraioli, google.com/+JuliaFerraioli
When am I ever going to use this?
... tales from a career in industry
Slide 2
Slide 2 text
Game Changers
A bit of background
Slide 3
Slide 3 text
3
@juliaferraioli
BA in CS @ Bryn Mawr
College
2003-2007
A career retrospective
2007-2009 2009-2010 2010-Now
Software Engineer @
Microsoft
MSCS @
Rochester
TE @
DocuSign
Cloud @ Google
Slide 4
Slide 4 text
4
@juliaferraioli
BA in CS @ Bryn Mawr
College
2003-2007
A career retrospective
Slide 5
Slide 5 text
5
● Machine learning with background knowledge
● Incorporating user interface to allow interactive
clustering
● Algorithm used a variety of improvement on good ol’
k-means
DREU @ UMBC
5
Humans are smart. Machines are smart.
Humans + machines are smarter.
http://goo.gl/yZkUjU
Slide 6
Slide 6 text
6
@juliaferraioli
research in your
very own AI lab on
developmental
robotics
Slide 7
Slide 7 text
Impossible is closer than we think.
Slide 8
Slide 8 text
8
@juliaferraioli
BA in CS @ Bryn Mawr
College
2003-2007
A career retrospective
2007-2009
Software Engineer
@ Microsoft
2007-2009
Software Engineer @
Microsoft
Slide 9
Slide 9 text
9
● Launched & shut down a product in 2 years
● Learned lots about what I didn’t want
● Demystified “big name” technology
● Used precisely none of my skills
Software Engineering
9
Slide 10
Slide 10 text
This isn’t for me.
Slide 11
Slide 11 text
Changing directions isn’t failure.
Slide 12
Slide 12 text
12
@juliaferraioli
BA in CS @ Bryn Mawr
College
2003-2007
A career retrospective
2007-2009 2009-2010
Software Engineer @
Microsoft
MSCS @
Rochester
Slide 13
Slide 13 text
13
@juliaferraioli 13
Slide 14
Slide 14 text
14
● New technology is made up of old technology
● Build for the world
● Applications always go further than you think
● (Reprise) People are smart
Breaking (down) research
14
http://www.vizwiz.org/
Slide 15
Slide 15 text
Everything is interconnected.
Slide 16
Slide 16 text
16
@juliaferraioli
BA in CS @ Bryn Mawr
College
2003-2007
A career retrospective
2007-2009 2009-2010 2010-Now
Software Engineer @
Microsoft
MSCS @
Rochester
TE @
DocuSign
Cloud @ Google
Slide 17
Slide 17 text
17
@juliaferraioli 17
Slide 18
Slide 18 text
Developer Relations @ Google
Slide 19
Slide 19 text
19
● Go back to the basics
● Those algorithms underpin everything
● Publish, publish, publish
● Always learn, always update
Research is everywhere
19
Use the research; don’t expose the
complexity
Slide 20
Slide 20 text
Game Changers
Research in industry
Slide 21
Slide 21 text
21
Cluster management
with Borg
http://goo.gl/1C4nuo
Slide 22
Slide 22 text
22
Cluster management
with Borg
the system we
internally call
http://goo.gl/1C4nuo
Slide 23
Slide 23 text
Game Changers
Image by Connie Zhou
Slide 24
Slide 24 text
24
@juliaferraioli 24
job hello_world = {
runtime = { cell = 'ic' } // Cell (cluster) to run in
binary = '.../hello_world_webserver' // Program to run
args = { port = '%port%' } // Command line parameters
requirements = { // Resource requirements
ram = 100M
disk = 100M
cpu = 0.1
}
replicas = 5 // Number of tasks
}
Developer View
10000
27
@juliaferraioli 27
Failures
task-eviction
rates and causes
Slide 28
Slide 28 text
28
@juliaferraioli 28
Failures
advanced bin-
packing
algorithms
experimental
placement of
production VM
workload, July 2014
stranded
resources
available
resources
one
machine
Slide 29
Slide 29 text
Game Changers
Slide 30
Slide 30 text
30
Having weird dreams
since 2015
http://goo.gl/OWjBWO
Slide 31
Slide 31 text
31
@juliaferraioli 31
What is this?
Slide 32
Slide 32 text
32
@juliaferraioli 32
What is this?
Possible answers:
● A painting by Seurat
● “A Sunday Afternoon on the Island of La
Grande Jatte”
● Example of pointillism
● People at a park
● Something that belongs in a museum
● A nice break from class
Very...human answers
Slide 33
Slide 33 text
33
@juliaferraioli 33
A machine’s view
Slide 34
Slide 34 text
34
@juliaferraioli 34
A machine’s view
Slide 35
Slide 35 text
35
@juliaferraioli 35
Deep learning * with neural networks
* very simplified
Slide 36
Slide 36 text
36
@juliaferraioli 36
Deep learning with neural networks
Slide 37
Slide 37 text
37
@juliaferraioli 37
Sometimes it doesn’t quite...work the
way we anticipate
Slide 38
Slide 38 text
38
@juliaferraioli 38
Possible applications?
Slide 39
Slide 39 text
39
@juliaferraioli 39
Are there non-technical benefits of research...
...besides weird pictures?
Slide 40
Slide 40 text
Game Changers
Impossible isn’t a limit; it’s a challenge.
Slide 41
Slide 41 text
Game Changers
A fine-tuned BS-detector is invaluable.
Slide 42
Slide 42 text
Game Changers
Complexity reduction applies to problems, too.
Slide 43
Slide 43 text
Game Changers
Critical thinking is a muscle.
Slide 44
Slide 44 text
Game Changers
Advice from me
Slide 45
Slide 45 text
45
Take all the opportunities you get…
...even if you think they’re weird.
Slide 46
Slide 46 text
46
...no matter how famous you think they are.
Talk to everyone...
Slide 47
Slide 47 text
47
...whether it be a language, how to blow
glass, or how to play an instrument.
Always look to learn something new...
Slide 48
Slide 48 text
48
@juliaferraioli 48
Slide 49
Slide 49 text
49
@juliaferraioli 49
Open source scientific
computing
● SciPy
● SciRuby
● Kubernetes
● Ruby Science Foundation
● Various universities
● Institute for Artificial Intelligence
Slide 50
Slide 50 text
Thanks!
Creative Commons image by Feans
Questions?