Most modern data science teams don't consist of the people we originally considered data scientists (IE graduate math, stats, ML, engineering (not CODING, ENGINEERING), and business acumen) and are thus disgusting sacks of garbage. Here's how to build yours!
1. Post a job requirement consisting of programming languages, frameworks, soft skills, and domain specific expertise. It is MUCH more important that the candidate have experience with (for example) hydrology than they know how to build scalable machine learning models. Make sure it's long and include obvious traits like "good communicator" and "attentive to deadlines".
2. When interviewing, ask the candidate to describe a project they worked on and quickly decide whether that sounds like something you will do. If it doesn't, move on.
3. Inform the candidate they will be using Spark. If they don't laugh, you're on the right track.
4. Have the candidate do a take home test. This is an excellent way to weed out people that don't want to waste their time, have kids, and / or have better options.
5. Low-ball the offer. You're hiring in one of the most competitive fields in the job market, the last thing you want is have a highly skilled data scientist feel your offer is fair.
6. Speaking of money, make sure the salary range is way below that of anyone with the word "manager" in their title. Managers manage people, which is obviously a much more complex, much more rare skill set than someone with graduate level knowledge of multiple fields. Good managers foster innovation right?
8. Once you've hired a small team, inform them of what you need and when you need it by. When they tell you science is about exploration and not all projects succeed, make sure they understand that at your company all science produces RESULTS.
