The One Technique to Land Your Dream Data Science Job

Individuals breaking into data science have diverse backgrounds. Many come from the computer science and programming side. Others come from STEM disciplines. Still others are (seemingly) anomalies like myself, coming from backgrounds such as the social sciences, English, journalism, and a multitude of other degrees/backgrounds. While there are various ways to break into data science, there is one thing you can do to stand out amongst potential candidates. But first, we’re going to review a study that will help facilitate in that process. 

WHAT HR & RECRUITERS LOOK FOR

In 2012, recruiters participated in a study that leveraged eye-tracking software to monitor their evaluation of potential employees. They were also asked to estimate how long they reviewed each individual resume. Self-reports indicated that, on average, they believed they spent 5 minutes reviewing each candidate.

Horrifyingly, this is inaccurate. To assess fit/non-fit, 

“Recruiters spend only 6 seconds reviewing an individual resume.”

Almost 80% of THAT time is spent in the following areas:

• Name
• Education
• Current title/company
• Current position start/end dates
• Previous title/company
• Previous position start/end dates

BECOME A DATA SCIENTIST

Thus, I generally recommend two courses of action: title-based roles or organization-focused roles. It can be potentially beneficial to accept a job with a ‘data scientist’ title regardless of the company itself. This is due, in large part, to the psychology of trust. Someone who has earned a data scientist title has been validated by the previous company and is therefore perceived as capable of accomplishing said duties.

The other option is to work for a renowned company. The organization itself brings clout. Imagine that you are a recruiter and have received two data analysts’ applications for a data scientist position. One works at Google — the other at a no-name startup. All things being equal, who do you have more confidence in hiring?

Now…what are the best paths to get a data scientist title? I’m currently conducting an analysis of data science courses and programs for those who aren’t pursuing graduate degrees. While this analysis will come in a later update, there is an approach that anyone can do to stand out.

Passion Projects. Capstone Projects.

They come in many names, but their primary purpose is to publicly show your capabilities to anyone interested. Oddly enough, it’s quite difficult to find good/thorough projects posted online. This means there is a tremendous opportunity to obtain a data science position for those who are willing to go above and beyond a simple Coursera course. If you don't have a Data Scientist title or work for a prestigious company, projects open doors. Many doors.

CHOOSING A PROJECT

When I mentor on Springboard for the Foundations of Data Science and Data Science Intensive courses, mentees often ask what dataset and project should they choose. I in turn tell them, 

“This is your opportunity to choose something YOU want to do. You don’t always get this opportunity working for organizations. Run with it.”

To help guide them I provide two recommended options:

1. Choose something that you are passionate about. You likely have domain knowledge in this space, which makes you incredibly adept at working with and manipulating the dataset (feature engineering). Moreover, you’re more likely to continuously think about the data and how to build better models.

For instance, I enjoy fantasy football. In the past, I’ve scraped football data and built models that did quite well on FanDuel and DraftKings. They did well because I know the space well and had ideas to improve the data for more accurate predictions.

2. The other option is to picture yourself 1-year or 5-years from now. What do you imagine yourself doing as a data scientist? Or what companies would you be enamored to work with? Reverse engineer your journey. Let’s say you want to work in a marketing department. You would benefit from building a churn model, attribution model, brand targeting models, or customer segmentation to present to potential recruiters. If you want to work in finance, find a dataset that allows for anomaly and outlier detection of customer fraud or build ARIMA models to project stock prices. There are plenty of options in this space.

PROJECT REQUIREMENTS

1. Find (or even better, scrape together) a dataset/s of interest
2. Explore/Initial Visualizations/Clean Data
3. Feature Engineering
4. Data Modeling
5. Model Optimization
6. Productionize: this can be an interactive website that allows your users to manipulate the data realtime e.g. RSHiny or an automated backend job e.g. Cron/Autosys/Oozie

EXAMPLES

Here are 5 capstones that capitalize on some of the aforementioned ideas:

MPAA Parental Advisory

Early Detection of Forest Fires

Practical Data Science

Predicting Cab Cancellations

Identifying zip code characteristics by income and education

CONCLUSION

Once you’re able to implement something along these lines, recruiters will consistently be knocking on your door. Good Luck!