Kima: You’ve hired many data scientists and data engineers; which role was the most challenging to fill, and how have you helped people be successful in these roles, especially with ML being at the core of that work?
Mike: There’s a tendency amongst us technologists, especially those of us who are really excited about machine learning technologies, to assume that ML is inherently valuable – like it has this intrinsic utility to artificial intelligence or this associated technology underneath that umbrella. The single biggest challenge is for leaders and hiring managers to recognize that the most impactful ML engineers, in particular, are those who don’t believe ML is inherently valuable.
It’s those who are focused on applied machine learning and will recognize that the technology is only valuable to the extent to which it replicates or scales some manual process in the physical world. It’s those who are able to connect the technology to the real-world impact and are first and foremost prioritizing the real-world impact that have the biggest impact on your organization. That’s the single biggest thing that we’ve learned and that I’ve personally observed.
That’s really important because there are some very intellectually stimulating and exciting applications of ML that are highly creative but don’t actually result in real-world impact and that that’s great. There’s a lot of value in those explorations. When it comes to moving the needle and transitioning your company from a SaaS 1.0 to SaaS 2.0, or from a company that goes from being scared of customer reactions to craving them, it requires a deep prioritization of the real-world impact that you’re looking to materialize. That’s something that can be coached, but it’s something hiring mangers who are looking to hire or data scientists and ML researchers should be aware of.
Kima: That’s interesting. What I’m hearing you say is that you hire a data engineer they somehow have high expectations of ML itself being the goal as opposed to being a tool that you use to solve a real business problem.
Mike: That’s exactly right, and what’s interesting is when you think business problem first or customer problem first, there are many ML techniques that are actually not the ideal solution. There are many regular expressions that would solve a problem more efficiently and better for the customer, and that’s that should be prioritized. It’s about solving the problem, not about the tool of choice.
Now, the ML toolkit is rapidly expanding, which is very exciting. But if you think tool first and not problem first, then you run the risk of not applying ML the right way and you run the risk of applying it too often.