Careers and data science.
Hi, I'm Dr. Werner Krebs, Ph.D., CEO of Acculation.
Acculation dot com.
[https://www.acculation.com] Hopefully, you'll subscribe
to our YouTube channel here and checkout one of these other videos on our channel
when you're done with this one, and if you like this video be sure to click
that like button.
So, what not to do when applying for a job.
A couple of approaches don't work.
"Hi, I'm a really bright guy.
Hire me!
I have a new idea!"
With the possible exception of strategy consultants, academics, and think
tanks, companies generally don't hire people just because they're smart.
They hire people because they have an existing business process need.
So, they're making widgets or making software and
they have to too many order for the widgets or
their software.
And so they look at Bob and they say,
"Okay, well, we've got too many orders coming in."
"So, what we want to do is: We want to take Bob over there and clone him."
Except, of course, human cloning is illegal.
So they take a step back and decide:
"Well, we can't clone Bob even though he's really good at making widgets."
"So, what else can we do?"
Well, they look at the job market and they say,
"Can we find someone who is just like Bob?"
"So, that we can meet our customer demand?" and
so along those lines: "Hi, I'm a really bright"
or, "Hi, I graduated summa cum laude in English!
You should hire me because I have this very impressive summa cum laude English degree!"
That doesn't work, right?
So, for most people that very impressive summa cum laude
degree in English means that the only thing they're really comfortable with is
writing.
And writing, unfortunately, by itself, does not pay very well because
there are a lot of other writers out there.
So, if you want to be hired just for being bright or for having a very
impressive college degree consider consider consulting.
And they'll usually hire you because you have an MBA or other application of
strategy consultant.
If you want to be hired because you have a new idea that
you think will revolutionize the market consider doing a start up.
Go through an accelerator or incubator.
Then, you'll be employed by investors backing your new idea.
And, I mentioned a couple of the other things that do employ people just because
they're very bright: academics, universities, and think-tanks
So, who am I?
I'm Werner G. Krebs I began programming at a very early age.
I was a University of Chicago undergraduate math
major worked for a Nobel Laureate in economics.
Did a Yale PhD in the this new field of bioinformatics, technically
Molecular Biophysics & Biochemistry, So
bioinformatics is computer science plus biology.
And this led to a career as a computer science or a data scientist worked for
some very famous people, did some time in academia, at a software company and then
on to finance as a senior analyst at what is now Bank of America
managing their billion hedging portfolio.
And, then, went on to the hedge fund Madison Tyler, now
known as Virtu Financial, of the top high frequency trading
firms in the world according to Wikipedia.
I led a team at a firm did marketing consulting for 90%
of Fortune 100 CMOs and learned marketing modeling from some of the
best people in the world in academic marketing.
Math and programming made all of this possible.
I managed a team of engineers and have now started my
own company Acculation dot com which does data science, artificial intelligence consulting
and software development among other things.
So, am a graduate of the Founders' Institute startup
accelerator.
This not easy: 60% of the people in my program didn't make it to
through.
So, it's been compared to survivor.
I have interviewed hundreds of people at different companies and hedge funds
and marketing and consulting group where I was hiring
manager.
I have been told in the past that over a hundred people were interviewed
for a position that i was eventually selected
for.
I'm going to point out again that the important thing to understand is that
businesses have business process.
They hire people to fill those business process needs.
They prefer people that can start right away, but what
they generally don't do is hire people simply because they're smart or have
a specific college degree without reference to those pre-existing
business needs.
I'll say a little bit more about now, and a little bit more at
the end.
So, I'm going to throw some books out at you.
"Thinking, Fast and Slow" is a 2011 book by Daniel Kahenman
Its central thesis is that there's an dichotomy between these two modes
of thinking: what he called System I are
the fast, instinctual, and emotional system.
This is the process that makes decisions when you're in a hurry, when you're being
chased by a lion.
It's about making decisions from the seat of your pants because
you're being chased by a lion.
And the is System II which is the more
slow, more deliberate, more logical system.
People do not behave as predicted by classical economics.
Those of you who were econ majors, you know were
brought up in this rational decision-making framework.
The truth is organizations tend to behave rationally
and there's a lot of evolutionary pressure on organizations: the tribe, the
family, corporations, governments to behave rationally.
Those organizations that do not behave rationally tend not to survive.
Evolution does not work on individuals.
At least for animals, It does not work on individuals.
It works on the species or it works on groups.
And, as a result, the evolutionary pressure to be
rational was not applied at the individual level.
There is actually a lot of evolutionary pressure for individuals
not to be rational.
That evolutionary pressure was applied at the group level:
on families, on tribes, and there's a lot of evidence that
individuals do not behave rationally.
Individuals are loss adverse.
They are more likely to act to avert a loss than
to achieve a gain.
Thinking slow: so, the scientific method is a form of thinking
deliberately to find the truth.
Is an outcome repeatable and reproducible, is it
option falsifiable?
It is the opposite of superstition.
The opposite of science is superstition.
Astrology which is one of the few religions that can be disproved.
These things that appeal to emotion or system one: that fast
way thinking.
Examples in political science are global warming and it's politically
motivated deniers.
Thinking slow by thinking fast at the same time: so, if you have the
resources --- you have employees and software like an Excel spreadsheet, you
have trainin, you have MBAs, you have the inclination: business processes and
incentive, survival, profit.
They try to think "slow" whenever possible if you're a company.
So you going to try to use computer models.
For example, Excel or spreadsheets for thinking
slow.
Business frequently try to formalize the most important and most
competitive processes to try to ensure quality.
Some examples are forms, procedures, assembly lines, quality
assurance inspection tests.
Ideally, businesses would like their most important, frequent
analyses to think both slow and fast: to think slow in a sense of thinking that is
deliberative and rational.
And, also fast, not because they're using System I, but
or thinking intuitively, but fast because the thing is being done
I computer model that can think very quickly.
Examples are high-frequency trading, marketing analytics.
So, another book for you!
"The Signal and the Noise" which is the 2012 bestseller by Nate
Silver.
The full title is: "The signal and the Noise:Why most prediction fail but
some don't."
And, it talks about building mathematical models.
The synopsis is to build a really good mathematical model you
really need to understand the field.
For example, in the baseball statistics in the
example he cites, knowing which parameters are important
and are reliable to select for in
that model, rather than relying purely on [automated] statistical tests.
And using a Bayesian approach to model building.
Another example is Moneyball both the book and the movie.
And, of course, Lewis the author also did some books on trading and
and high frequency trading, where mention some people I've worked with.
Big data: so another book: "Big Data: the revolution that will transform
how we live, work, and think" came out in 2013.
It talks about Big Data being a consequence of Kryder's Law, which is Moore's
Law for disk space.
We can now store, slice, and analyze complete data sets.
For example, all of Amazon purchases.
We have enough data to build the statistical models
on some obscure phenomenon, something that was not
previously possible.
It's only become possible in the last two years.
Previously, we needed to subsample data to build statistical models.
Now, we can program a computer to go through all of
that data, build it's own statistical models, and
try to discover correlation and causation.
For example, Google discovering certain queries where highly correlated to Swine
Flu Virus.
Modeling and political Science.
So example, of a business is Robert Pape's 2005 book: "Dying to Win:
the strategic logic of suicide terrorism."
Robert Pape from the University of Chicago.
Most suicidal terrorists,
according to his research, are altruistic, well-educated, nationalist, motivated, and
they're fully witting, and dedicated to their fatal mission as a service to their
community.
So, this is an example of where you can you use big data to discover
things and use that to try to learn things about what motivates people and
how you can potentially discourage this.
Another example in political science is something I was a little bit involved
with, which is Prof. Heckman's work on the Job
Training Partnership Act (JTPA).
And, this was used to tweak legislated eligibility requirements to try
to maximize access to the intended audience, which in this case with the the truly
unemployed needing skills, and try to minimize free-riders.
Some of the potential free riders were seasonally
unemployed teachers who might appear eligible because of that summer break, and
then people who are potentially quitting their jobs try to meet
the requirements for getting this potentially very valuable training.
By using data they were able to tweak the
legislation so that seasonally unemployed teachers would not be eligible by
requiring the person the unemployed for more than three months and putting in certain
other things to try to ensure that the people who are applying really needed the
training.
So people ask me, "What should you know to get a career in data?" at a
minimum possible you should try to learn SQL and things that are SQL-like
as well as NoSQL databases these days.
Now, people ask, "What if you only want to learn
one programming language?
What should it be?"
The lowest common denominator for analysts
these days is VBA.
This is the programming language that is built into
Excel.
So, if you own a copy of Excel spreadsheet, there's a programming
language built in, and it's called Visual Basic
for Applications [VBA].
Every analyst at a Wall Street firm at a minimum is expected to
know that programming language.
If there is just one programming language that
you put on your resume, this should be it.
Moving up from VBA, you've got Java, Python, Ruby, C++, others like C#, and
visual basic dot net [VB.net].
Visual basic dot net is something very different from VBA, although
they're both Microsoft products.
Perl, statistical programming, so R and
everything else: SAS, Stata, SPSS, Eviews, Gretl.
Mathematical programming: Numerical Python [numpy], SAGE, Matlab, Octave
Mathematica, Maple.
In the last two years we've also seen Scala become important
in part because it's the preferred way of interacting with the Apache Spark
toolkit for doing data science in the cloud.
Engineering software and hardware, hacking for speed reliability, and then
software tricks: algorithms, compilers, network latency optimization, cloud, non-
cloud, and then hardware tricks: even implementing algorithms in hardware
like Field Programmable Arrays [FPA].
Modeling methods: you've got regression and modifications, classical
statistical techniques like linear regression are still very important.
These can be partially automated through
things like statistical stepwise significant testing, although this can
violate statistical assumptions.
You've got segmented regressions, manual partition decision
trees, low-data regressions, waited average multi-models, nonlinear regression transforms,
log-log, automated shape discovery with software that tries to figure out what type
of transform to apply (for example a log-log transform)
to make linear regression effective.
What is the advantage of linear regression?
Well, it's computationally very easy and this becomes very important when dealing with
large amounts of data.
Then there's the new stuff right?
Machine learning: this tends to require a lot more data, computation
is slower, and it obscures the parameter meaning.
You've got Hidden Markov Models [HMMs], fuzzy logic, neural
networks, genetic algorithm, Bayesian networks, and then manually constructed
expert systems, which are low data, high-expertise, and tend to
be Bayesian or manual decision trees.
Clustering techniques, decision-tree related related, you've also got self-organizing
maps and parameter selection through Principal Component Analysis [PCA].
techniques, and of course, in the last few years, we also have Deep Learning.
There's the Pareto Rule of Efficiency. were twenty percent of your efforts gets you
80% of your results.
You can use all of these relatively simple and transparent
techniques that I've just described, like classical statistical regression, fuzzy logic,
symbolic artificial intelligence, to come up with a first model, often a very
fast model that will get you 80 percent of your accuracy for twenty percent of
your effort and then you can use Deep Learning to use a neural network
to correct those errors in that models.
So that twenty percent of time when that simple first step is wrong can be
corrected by the neural network.
This is a very powerful technique based on the way to
human brain works.
It's been proven very effective.
Getting a job on Wall Street . So, I already mentioned SQL and Excel VBA
typically going to be minimum requirement for analysts.
If you put down only two skills on your resume put down
SQL and Excel VBA, and, if [you're limited to] only one programming
skill, I would say Excel VBA.
Know those well, because you will be asked questions
about them.
Some mathematics and finance background, a math/econ/stat degree
CFA candidate, MBA, quantitative PhD are all very helpful as are hard programming
skills like Java, C/C++, Python, Scala, in general, are all useful.
Don't lie on your resume,
don't put programming languages that you've never used, don't claim you can
program when you can't.
You can programming from books, online videos, websites.
What's employable: MBAs JDs, MDs,, and PhDs in most quantitative fields
tend to get taken a lot more seriously than people who just have Bachelors [degrees].
People who just have Bachelors are taken a lot
more seriously than people who just have high school.
And people would have actually graduated from high school are
taken a lot more seriously than people that have gotten the GED degree.
like JD/MBAs, PhD/JD, PhD/MBAs, and PhD/MD MD/MBA tend to get taken a lot more seriously.
MD and JDs are licensed professions: the degree
you get tends to be a lot less important than having the license.
The degree is required for the license.
This is in contrast to MBA.
Salaries for MBAs tend to be highly variable.
Stanford MBA graduates typically make as much 120K a year
on average at graduation, whereas
someone from a lower tier MBA school might only make on average 40,000.
(This is as of a few years ago.)
Most of the top-earning MBAs a few years back went to Wall Street.
Today it might be more consulting.
The low paying MBAs tend to be found at non-profits.
Historically, military veterans with an honorable
discharge and demonstrated technical skills were taken more seriously than
high school graduates and in some cases more seriously than bottom tier college
graduates Military officers often became CEOs especially after business
school.
This may be changing.
Starbucks pointed out that military veterans have twice
the unemployment rate of the general population.
This might be because of the psychological trauma of warfare and
something present-day that wasn't seen in military veterans two decades ago.
Quantitative and programming skills are always in demand.
They get taken very seriously.
Working at famous companies, institutions, and [with famous] people, especially
in positions that are highly selective and
hard to attain.
You can follow Warren Buffett's advice of ignoring the salary
and working where you will learn the most.
So working for some famous person in some field, even if that
doesn't pay as much as well as working for someone who's not famous
was not as educational of an experience, but pays
well.
What if you can't program [and] you can't do
quantitative?
Well, an MBA is the good as it doesn't require as much quantitative [skill].
Sales and marketing skills can also be very
employable.
These are people skills similar to being an actor.
Some people call them "expressives."
These can be very high risk.
Although some salespeople make six-figure salaries, unlike programmers there
are a lot of people making very low salaries of 20,000 or less, and salespeople
are in one of the few occupations that are potentially exempt from minimum wage.
So you have some salespeople that are making that are purely on commission
making a salary of 0 and this is not uncommon in commercial real estate, for
example, where a person might be very happy to do one big deal a year, and,
until/when they make that deal they'll get a 40 thousand dollar
commission, but don't spend most of the year trying to make that deal on a salary of 0.
It's very high stress as you imagine.
Related to sales and often requiring an MBA is business development.
So, this is a sales person who tries to negotiate
deals between startups and potential partners.
This is a good person for an engineer to partner with if you want to do a startup.
So, yeah, what not to do when applying for
a job.
"Hi there, I'm a really bright guy!
Hire me because I'm really smart!"
Or, "Hire me I have a new idea!
Hire me, I graduated summa cum laude in English!"
Why?
We've already gone over why these approaches
aren't that good, right?
Companies look for people that are similar people to people they've already
hired.
They look for Bob the Widget Maker and they want someone similar to Bob the Widget
Maker.
So let me throw another book out there for you it's called "The Black Swan."
This is the nonfiction bestseller by Taleb, not
the fictional movie with Natalie Portman.
So, in "The Black Swan," Taleb talks about why professions have different risk levels.
The extreme value distribution is
counter-intuitive to the human psyche.
Many "media" professions like acting, writing generally don't make as much, because
they have an extreme value distribution in their salary, which is counter-intuitive.
A tiny fraction of the very best dominate the
media and they tend to make huge salaries while the rest starve.
So if you think you're a good actor you might make
a good salesperson.
They have similar personalities and actors hock a lot of
merchandise.
Salespeople, however, tend to have a less variability in
their salaries than actors Non-"media" professionals like lawyers and doctors
tend to have even more even salaries.
Unlike an actor
who typically makes very little, doctors, lawyers, and programmers tend to do very
well.
While the best doctors don't make that much more than the average doctor, and
the best doctors make a lot less than the very best actors, even though the
average actor makes a lot less than the average doctor.
So, for every one JK Rowling billionaire author, there are
hundreds or thousands -- maybe millions of starving writers out there.
If you want to make it as an actor or a writer
it helps to get into the best acting and writing programs, because these give you
credibility and name recognition of the top of the field and visibility to people
that can actually hire you into the most lucrative positions.
Startup accelerators: if you're adventurous and want to get funding for
your company you can apply to accelerator.
The terms accelerator and incubators technically mean different things
but in practice are used interchangeably.
The most famous are Y Combinator TechStars, and Founders Institute.
But there are many others out there.
Several out here locally [in Los Angeles] are Amplify
LaunchPad and Science.
It helps to have a high IQ or have attended an Ivy League
University.
It's almost a requirement at some of these startup incubators.
Knowing programming or marketing sales, or -management is also very helpful.
This [doing a startup] is very high-risk.
Ninety percent of startups fail
according to some statistics and in contrast Y-combinator historically
(this may be changing) but historically Y-combinator would accept less than 1%
of its applicants, while 90% of its program typically went on to achieve
multi-million dollar valuations.
So, getting into Y-combinator usually changes the
risk profile: you go from something where the odds are you're
going, to fail something where the odds are that you're very likely to succeed.
That's why these are potentially useful.
internet of things.
A few years back a $35 computer came out.
The Raspberry Pi It ran Linux, it's connected to hdmi TV, it
had ethernet, a USB for keyboard, and you can
even have WiFi.
It used an SD card for its hard drive and an old smartphone charger
for its power supply.What is the take home lesson?
Well, if computers cost only thirty-five dollars, they are going to be
in everything.
Everything, everywhere.
My bio again: programming at an early age,
Yale PhD in bioinformatics-related or a data science-related field
and sometime, in academia, a software company, and then a lot of finance.
Was a senior analyst at a major bank,
worked at a major hedge fund, did work for a leading marketing firm that
later went on to become NYSE- listed and have now started my own company
which does data science and artificial intelligence consulting among
other things.
Thank you so much for your time.
Be sure to hit that subscribe button.
If you like this video, or if you found this video
enjoyable, I hope you consider liking us.
Check out some of our other videos.
Of course, visit us at our website at https://www.acculation.com .
Links should be appearing in this video as well to our web site.
Thank you and have a great day.
For more infomation >> La démocratie permanente - Comment ça marche - Duration: 2:58. 
For more infomation >> 2017 Lexus® LC Commercial
For more infomation >> Kat Von D Lock It Podkład/Baza + tańszy identyczny! Zamiennik Hot or Not??? - Duration: 23:35. 
For more infomation >> CVLF - Salade de chou kale - Duration: 4:03. 
No comments:
Post a Comment