Hiring and Building a Data Science Team

Problem

"When you have decided to build a data science team you want to begin thinking broadly. You first must identify your bottleneck, or constraint that you are looking to overcome. This includes not only today's constraint but also thinking ahead to the next one as well, which is not as obvious. Once identified, then you need to know how to start sorting through people in order to hire the appropriate people onto the new data team. Below are a few key categories to contemplate when hiring, and my advice on how to sort through them."

Actions taken

Specialist vs Generalist

"Someone becomes a specialist because they're interested in knowing one thing. If you are looking to solve problems in that particular area over and over again then hiring a specialist would be the way to go. However, if you have an unending number of problems then you will probably want to look for a generalist. Especially if you are just starting to build a data science team, hire a generalist who has coding skills and can apply whatever tools are necessary. They will be able to move quickly and actually work with the data. Of course, it's best if you can also find someone with business skills who can help solve the larger problems."

Data Scientist vs Data Engineer

"The title and role of a data scientist rests at the top of the pyramid. In order to get there you have to start from the bottom and work your way up through all the rest of the layers. Thus, if you are just starting a data science team it would be best to start with one of the middle layers categorized as data engineer. This is due to the fact that what you usually need is less sophisticated algorithms, easy access to development data, and solid pipelines. You need to start off with simple models and then you can progress to more sophisticated models. Therefore, hire a data engineer. A data scientist hired into this role would likely become irritated that all they were doing was simple modeling; problems would not be well-placed."

Education/Experience vs Passion

"I would always go for the person with passion. They don't necessarily have to understand all of the code, but they do have to have an understanding of certain aspects. In my opinion, it's better to hire someone who is smart, cares, and is willing to try hard over someone who has a PHD. You simply can't extract passion out of people."

Lessons learned

• "There is a range of people who classify themselves as data engineers. For example, a bad analyst who started writing bad pipe code and thinks he's an engineer. Or a high-fly infrastructure engineer that is writing the next version of Flink. These two extremes and everything in between are considered data engineers, so be very clear in the beginning about what you are looking for."
• "Pairing is a great way to grow your data science team. Take somebody who is a good solid engineer and wants to learn this kind of material and pair them with a data engineer or analytics person. They can learn from each other and learn together how to build more robust pipelines and infrastructure."
• "Getting value out of data science continues to be very difficult for almost all businesses. This is often because of a mismatch between expectations, roles, and people. 'Data Science' is a very broad term these days, so think carefully about hiring generalist data pros, rather than e.g. deep learning specialists."