7 Tips to Help Attract and Retain Data Science Talent

Day by day, data is proving to be the lifeblood of industry. With the explosion of analytical tools, open source programming languages, AI and machine learning, the transformative power of this asset is clearly driving a new kind of revolution in the digital age. 

It’s a good time for us data scientists, who apparently hold the ‘sexiest job of the 21st century‘ (thanks, Harvard Business Review). But while large organizations like big banks have deep coffers to snap up top talent, throwing money at the problem isn’t a long-term solution. For every job that might be displaced by automation, nearly two will be needed at the intersection of human, machine and algorithm.

Reskilling and upskilling should be on every organization’s agenda. Even with university enrollments in computer science tripling in the last 15 years, there’s still a huge shortfall of skills, and this will grow exponentially as new tech and programming languages emerge.

Getting us is one thing. Keeping us is an entirely different matter. Here are my recommendations to help find, and more importantly keep, data science talent.

1. Stop chasing the Harvard PhDs

Getting the best team doesn’t mean you need to staff up exclusively on PhDs. Curiosity fuels our work, so look beyond data science or math degrees and seek out people who are all about the learning. In our open source world, you could find a solution on Reddit, so your most powerful weapon is a team of inquisitive minds. When I’m hiring, I look beyond pure academics for innovative thinking and a willingness to learn. This creates a more diverse and successful team.

2. Don’t think we’re all just like you

Just like any team, we’re motivated by different things. An ML engineer or data scientist might not see success in the way that you do. They might be more comfortable hanging out with their code and data than getting on stage in front of the entire company to talk about a breakthrough they made. Data or business analysts tend to be more comfortable working with organizational stakeholders, as they tend to have a blend of technical and communication skills that bridge the two worlds. Understand the nuances of the roles and don’t project your commercial culture too heavily onto us.

3.  A little understanding goes a long way, so get involved

We want to succeed as much as you do, but we need data. Bringing us a handful of records or dumping a bunch of dirty data is frustrating for both of us. Get a window into our world by taking part in a hackathon that your company organizes or take a short online intro course on data science and AI. An hour’s investment really helps separate fact from fiction when it comes to AI and data project feasibility.

4. We gravitate toward each other

We know that you likely see us as solitary animals. But honestly, we are a creative bunch that loves to solve problems together. But more than that, we’re competitive, too. If one of us is stuck on something, they’ll tell the team and it’ll be a race to solve it in the most creative way. Time complexity is our drug. So, don’t keep us away from each other for too long -- we’re pack creatures at heart.

5. Time and space work differently for us

It’s no secret that the 9-5 just doesn’t work for us. Sometimes, we’ll be working all through the night to get to an endpoint, and equally we can’t just flip a switch at 9 a.m. to solve a problem. Sometimes, too, we need to spread our wings to tackle a problem. So, to keep your team happy, let them wander to AI meetups or community events. Some of us have greater bonds with the research community than the commercial one, and our ‘watercooler’ might be a side project. So, don’t confine us to set hours -- let us wander free occasionally.

6. We need the freedom to fail, and the tools to experiment

We’re a curious lot. We need our skills to be used and to be cognitively challenged. The fail-fast paradigm works well for us: We iterate until it works, and we’ll probably have multiple projects running at once from our horizon model stage 1, and potentially 2 and 3. Because of this, we need a robust infrastructure, whether we’re working in a dev environment or experimentation in production.

7. Share the success

Let us know that our work is valued. Many projects don’t start with a feasibility model, and the failure rate is high. So, when things work, let us know about the impact on the business. We may sometimes act like all we care about is models and outcomes but knowing that we contributed to a successful commercial outcome makes us happy, too. And don’t forget to give us data back, we need to continuously retrain our models to make them better.

Adam Lieberman is the data science lead at Finastra, one of the largest providers of financial technology in the world. He works across all financial lines of business, from lending to retail to treasury and capital markets, and leads the technical development of artificial intelligence and machine learning based solutions.