Where to Find Data Scientists: Train from Within

At the Data Science Association, we frequently receive inquiries from companies looking to hire data scientists. There aren't any. OK, even if you can find one:

  1. They won't have the domain knowledge of your company.
  2. They'll be very expensive.

The alternative is to train from within. With this approach, you won't have to train on your domain knowledge (which most companies underestimate because they take such knowledge for granted as they live and breathe it day in and day out). So you're trading one type of training for another.

If an internal employee already has math skills and aptitude as well as conventional IT skills, then training on Big Data, Machine Learning and Data Science shouldn't take more than six months. In contrast, learning a company's domain knowledge can take 12-18 months, even though managers and HR may estimate it at taking one month.

The above diagram is, of course, based on Drew Conway's original Venn Diagram.

The problem is especially acute here in Colorado. Colorado companies are notorious for not paying for relocation, yet recruiters from the West Coast regularly making poaching trips through Colorado. The result is a net brain drain and recruiters repeatedly soliciting the little remaining talent.

Short of paying outrageous salaries, companies would be better served to move a current mathematically-talented employee off to learn and do data science, and then back-fill that employee's position with a conventionally-skilled and more affordable hire.

UPDATE 2015-01-22: The diagram at the top has been updated. The old diagram is here.