How To Land Your First REAL Data Science Job

I had a surprisingly high number of followers ask for insight on this topic so here is a brain pgdump. This is a weird topic for me, it's like someone asking me how to snowboard. My response would be "I don't know, you just do it". Nobody was around to show me or my data wrangling buddies how to become a data scientist. They just did it. When I asked them for additional insight on this post topic they just laughed (arrogant jerks).

Data Scientist: The Sexiest Job of the 21st Century
- HBR, 2012

In 2012 the Harvard Business Review released an article claiming that data science would be the sexiest job of the 21st century. This insight was due to the limited supply to the market and the exponential growth in demand. Since then the market has seen an oversupply of some really bad talent mixed with sporadic gems.

Everyone who was previously an analyst, statistician, or engineer has self-branded themselves as a data scientist for good reason. They make more, in some cases a lot more.

On glassdoor.com the national average salary for an analyst is: $66,565, the national average salary for a data scientist is: $118,709. If you look at the upper end the gap is much more pronounced. Hence the flood of unqualified applicants.

For those readers that want to land a legitimate data science job and realize the sexiest job of the 21st century here are my 6 tips to help realize that goal.

1. Merit Trumps Degree:
Good news job-jumpers! Degree is not a hard constraint. Some data shops will want a graduate degree for their consultants, however that is not required for many data science jobs including data science jobs at HireVue. The data speaks for itself, pun intended, so if you have plenty of merit in your abilities you shouldn't worry about having to go back to school. On the flip-side taking an online self-study course like https://www.coursera.org/course/datasci doesn't make you a data scientist either. You need to do a lot more. Don't walk around the topics, jump in head first and drown yourself.

2. You MUST program well:
You need to be a solid programmer. You should try to be better than any one you know that still calls themselves an analyst, statistician, or engineer. If you don't have a strong programming background no worries, you can try a python bootcamp, or you can make it a goal to start programming for an hour a day using YouTube and blog tutorials.

3. You MUST "math" well:
The best data scientists are not constrained by the book. They can alter/tweak/modify existing algorithms to work as needed for their specific problems. This skill set can be challenging, but is also not impossible to teach yourself. There is also a boot camp for this skill set HERE. I would recommend going through all of the python sklearn examples (supervised learning / unsupervised learning, etc..). Then once you are familiar with what the algorithms do begin reverse engineering them at the math level, try to write some of the basic ones from scratch (i.e. bayesian methods, regressions). What is the objective? What is the difference between OLS and PLS? SVM? PCA? What are eigenvalues and how do I calculate them? Why are they important? What is PINV and how is it different than INV? What is convolution and why is it so fast for moving average calculations? A great way to jump start your linear algebra is with image processing. It is fun, you see visual eye candy along the way, and linear algebra will help you do it well. Image processing is also a good skill set to have anyway for feature extraction so go ahead become an expert and add it to your toolkit. Here is a great resource for getting into python image processing, check out the author's book as well.

4. You MUST be involved in social media or your local community:
The data science realm changes very quickly. You will miss the train. I would guesstimate that we are moving at a 1 year pace right now that if you are out of the loop for a year you will miss enhancements in algorithms, libraries, training sets, or projects that can accelerate your abilities. Nobody wants to hire someone who will become stale on the shelf in a few years so strong community engagement is attractive to employers. Keep your ears to the tracks nerds and show your community involvement. Also networking in your community is a great way to land your first data science job by allowing recruiters and hiring nerds the ability to get to know you.

5. Make yourself uncomfortable on a regular basis:


Think you are smart because you know k-means, random-forest, genetic algorithms, grid search, etc... well so do all of the other self-branded applicants you are competing against for the job. So how are you going to stand out? One tip I enjoy is scheduling myself for a talk or presentation on a difficult topic I know little about. Once you are scheduled you will do what it takes to become an expert before you give the talk to your local MEETUP or conference. Think you could use deep learning on high frequency bitcoin algo trading? Does it intimidate you? Does it sound fun? Good. Go sign up for a talk and spend the next month sweating while you figure it all out. So what does this do? It builds confidence. It improves your communication skills, and being a speaker increases your visibility in the community. Employers want to hire someone who is confident and knows what they are talking about. Teaching a technical topic and answering questions from an audience will sharpen your skills.

6. Purple-squirrel-breadth trumps depth:

A common term with data science is the idea of the purple squirrel (the data science who is an expert at everything). Obviously we all know nobody is, however for the interview you need to be. If you are interviewing for a python shop and you have 10 years of R experience, the other candidate with more python experience will have a leg up on you vice versa. So find out what the technologies are from the job description or social profiles of the data nerds there and make sure you have spun up solid examples on those technologies before the interview. When you interview you will have kinks in your armor (technologies you haven't tried, infrastructure you aren't familiar with, etc..), figure out what matters to that employer and fix your weak spots before the interview. When decision makers are splitting hairs between the top candidates that could make the difference between you and the next applicant.

7. Communication:
What does communication have to do with data science? Everything. You must be able to impress your interviewer with your ability to communicate complex topics to a general audience whether that is the customer or your boss. Frequent public speaking will help improve this. I have seen strong technical data science interviews fail to continue because of doubts around the candidate's ability to communicate effectively with customers.

8. Get an AWS account:
If you don't have an AWS account yet get one! Must do. Budget $20-50/month to play on the cloud weekly. Spin up instances, build clusters, stand up ipython notebooks, web servers, databases, EMR, etc... Applicants with real cloud experience will be much more convincing with experience with using all of the new technologies (sparkling water, etc..). Play with SALT, Starcluster, Apache Spark, EMR, S3, LAMP, etc.. How about you build an app this weekend to allow users to upload zip files of good/bad photos that you automatically do deep learning on. That would be a great exercise. Next weekend build out a spot instance infrastructure to reduce your costs. The weekend after that port it to Amazon GPU instances, etc...

9. Linux = good:
The terminal is your friend. The more you play in the terminal (mac/linux) the better off you will be. Sorry Windows users, the windows power shell isn't going to cut it. Stuck on windows? There is always cygwin, putty, and working remotely on AWS on linux instances. Here is a resource: http://datascienceatthecommandline.com/

10. Fall in love:
Your future employer can sense passion. If you LOVE data science that will come across in the interview. Find the topics, meetups, and projects that make you LOVE what you do and you will be unstoppable.

Quick pet-peeve:
If you are in the market for a data science position DON'T change your linkedIn status to "looking, available, hoping for a data science opportunity". Those make you seem like a noob who needs their hand held. Instead just fight hard to be epic and become more active with your face-to-face networking to find your job/internship. If you know where you want to intern reach out to someone at the company and just ask them point blank "What can I do to impress you so much that you would consider me for an internship?"

If you are a successful data scientist and you have other insights or feedback please comment below so I can update the article to reflect more of a median opinion.

-----------------------------------------------------------------------------------------------------

These are 4 of my most popular articles:

This Is Why Your Data Scientist sucks





4 reasons to work for a startup instead





Death of the data scientist





This is why your plots suck

Keywords:
big data, hadoop, spark, random forest, support vector machines, supervised learning, clustering, data science, data science jobs, cloud computing.

Venkatesh Khanna

Business Analyst / Consultant, Six Sigma Black Belt, Independent Trader & Investor

7y

Very insightful about, (a) what more I need to work upon & learn (b) a status message I wont change! (an Aspiring Data Scientist :) )

Alfonso Then

EMBA | MMath | BEc • Let's create value through connections (people, data, insights)

7y

Learning by doing, by testing, by evolving, by experience, by loving it! Really good and motivating article ... Thanks for sharing!

Hatem Kotb

Product Manager, Data Analyst & Systems Coordinator @ WFP | Tableau Developer | Certified Professional Trainer

8y

Informative. Thank you :)

Like
Reply
Haardik S.

Business Intelligence I Data Engineering | AWS Certified

8y

Thanks for writing this informative article Mr. Benjamin Taylor. This provided many inputs which I can definitely relate with, especially about extracting features from Image Processing and reusing that mathematical concepts in data analysis.

Brian Schermerhorn

Senior Frontend Engineer @ Meta

8y

Ben, Do you have a more detailed opinion of the quality of the instruction/environment of Zipfian (now owned by Galvanize)?

Like
Reply

To view or add a comment, sign in

Insights from the community

Others also viewed

Explore topics