Return to Technical Hiring Guide Data Scientist Data scientist is not a new role; however, with computing resources becoming more affordable and technology advancing more rapidly, more and more organizations are employing people in these roles to help them understand their data. A major goal of data science is to make it easier for others to discern meaningful conclusions from data sets. And depending on the organization, the ‘customers’ can change. Sometimes data scientists are responsible for R&D, trying to research new business opportunities. Other times they can be a part of marketing teams (usually called “analysts”) and are responsible for providing insights into strategy, such as product positioning or segmenting existing customers. This role can require a versatile skillset, depending on the needs of the organization. Some roles require a thorough understanding of the challenges of processing very large data sets (i.e. "Big Data”). Others require practical applications of exploratory data analysis, statistics, machine learning, and data visualization. Data scientists may have computer science backgrounds, but they can also come from math or analyst backgrounds. When hiring for this role, be sure to understand the amount of programming required (and in what languages), since coding abilities can vary among candidates, and may not always be necessary to be successful. Data scientists can also be more specialized, depending on the needs of the organization or a candidate’s background. For example: data miners specialize in mining data looking for patterns or insights, and machine learning specialists focus on building and training models for repeatable analysis like artificial intelligence. Questions for Data Scientists:
- Q: Tell me a success story from one of your data analysis projects. A: Try to understand what made the project a success. Ask about the size of the data set, and how long they had to analyze it. What were the results, and how were they used? Then, ask about a project that was less successful and the surrounding details. .
- Q: How do you know when your results are good enough? A: One of the challenges with data science is that there isn’t necessarily a right answer. Sometimes there are lots of ways to solve a problem, and new algorithms and approaches can make results slightly better. However, it is hard to verify whether those results are correct, so ask for examples of how to verify correctness in their findings. Some strategies involve creating golden sets of data that you know are definitely correct and testing your algorithm against those, or sampling some of the data and verifying those results are correct. .
- Q: What types of clients or customers have you worked with in the past? How long did your projects normally take? A: Because these roles can vary so much, make sure the nature of their experience translates well to the position. You can also ask about working on a team and collaborating to share their results with others.