THE OPEN SOURCE DATA SCIENCE MASTERS (THE DIY DATA SCIENTIST) Clare Corthell Data Scientist at Mattermark @clarecorthell www.datasciencemasters.org
Jul 08, 2015
THE OPEN SOURCE DATA SCIENCE MASTERS(THE DIY DATA SCIENTIST)
Clare CorthellData Scientist at Mattermark
@clarecorthellwww.datasciencemasters.org
Deal Intelligence Platforminterface to live data about private companies
TODAY
• What a Data Scientist does
• Paths to becoming a Data Scientist
• Where to start
• Navigating a path
• Why you should run toward hard things
WHAT DOES A DATA SCIENTIST DO?
Data Scientists turn data into knowledgeby answering the right questions
Which is also predicated on asking the right questions
HOW DO I BECOME A DATA SCIENTIST?the answer you don’t want…
There’s no paved road, no one way
PATHS
1. Get a Classic Masters from an accredited University <Warning> I have yet to see one that’s better than the OSDSM
2. Attend a Bootcamp or Academy• Zipfian Academy (SF)• Insight Data Science Fellows (Palo Alto, NYC)• Data Science Retreat (Berlin)
3. Self-Taught• The Open Source Data Science Masters
THEORY & APPLICATIONor, why universities haven’t figured this out yet
Universities don’t focus on “Data Science” because it’s tightly bound to application.
Universities develop theory. Businesses develop applications.
The two exist symbiotically - they do need each other.
The goals are simply very different.
• Math• Computing
• Algorithms• Distributed Computing• Databases• Data Mining• Machine Learning• Graph Theory• Natural Language Processing
• Analysis• Visualization
• Python (language & libraries)
The Open Source Data Science
Mastersbit.ly/dsmasters
The internet helps me curate -
hence Open Source
(that’s alot)
CLARE’S PATHPreviously Product Designer, front end dev
Transcript bit.ly/corthelldata
6 months of study
Data Scientist & Machine Learning Developer at Mattermark
My team builds domain-specific systems for classification, recommendation, prediction,crawling, fact extraction, and more
languagesPython
SQL
machine learningScikit Learn
data manipulationPandas Numpy
matplotlib NLTK
designhtml/css/js
1. Get a goal2. Get a plan3. Get mentorship4. Get a project
1. Get a goal
What kind of “Data Scientist” do you want to be?
Explore the different roles
Pick something that sparks your interest
Find out what those people do on a daily basis
Rachel Schutt, Doing Data Science
Analyzing the Analyzers, O’Reilly
2. Get a plan
Figure out what skills you need to be minimally effective
Design a Curriculum (fork the OSDSM!)
Plan a schedule of study
Dave HoltzAirbnb
3. Get mentorship
Talk to people on twitter
Ask to buy them coffee (with a specific need or question in hand)
Get informational interviews(a lost art; they can turn into real interviews, but are low-pressure)
4. Get a question
Project Use real-world data to answer a question Who do iguana owners connect to on twitter?
Work on a real business problem Help a non-profit* with data they don’t understand
What channels of marketing are working for us?
*Orgs that coordinate working with NGOs: Bayes Impact, DataKind
(make it a small question - don’t set yourself up for failure)
Let’s talk about where this perfect plan gets really incredibly difficult
(Let’s start with a tautology)
HARD THINGS ARE HARD
Hard things are hard because there are no easy answers or recipes.
They are hard because your emotions are at odds with your logic.
They are hard because you don’t know the answer and you cannot
ask for help without showing weakness.
Ben HorowitzThe Hard Thing about Hard Things
When something scares you run like hell right into it.
The hardest things are things people avoid the most.That’s your marginal advantage.
Maybe that’s why there aren’t enough Data Scientists.
You will figure it out. It’s about ego management and problem solving.
RUN TOWARD HARD THINGSChoosing what you want to do
and what to work on
Not knowing everything
Being overwhelmed
Time Management
Math
Coding
Not knowing everything Being overwhelmed
There are a million things you could learn and work on. That’s overwhelming. But you can’t afford to get overwhelmed.
You won’t know everything. It’s impractical and impossible to know everything.
Learn to say “I don’t know.”
FYI Programmers don’t read books. They reference them as needed.
Time Management
How do I do all of this in a reasonable amount of time?- You don’t.- Be rigorous.
Ask yourself: Will this directly help me achieve my goal?
Refine your goals, focus your work.
Don’t switch tasks. Focus on one thing at a time.
Why is time management so hard?
We’re used to other people telling us what to do;
TeachersManagersParents
CODING IS HARD.
a hint for those new to programming
stackoverflow + problem
why code?
HUMANS SHOULD BE HUMANSAND
COMPUTERS SHOULD BE COMPUTERS.You must code.
Because automation.And no, there is no shortcut.
YOUR ADVANTAGE
Self-study in Data Science is hard.
But what you spend in energy and commitment to self-teaching is returned to you in:
• Choice of professional focus • Respect from potential employers for managing yourself. You
want to work with people who will respect and recognize that.• Skills that are tough to get from a university or employer• A path with no gatekeepers - no one will stop you.
Take the first step.
1. Learn to code in Python.2. Take Intro to Data Science (UW)3. Go get a coffee4. Ask one question
i ♥ questionsdatasciencemasters.org
@clarecorthell