4 Ways to Empower Data Science Teams

Author: Laura Ludwig, Practice Director of Data Analysis & Visualization


Perhaps you can relate: as a leader in data who works with data scientists, I’ve prioritized building a diverse team, in both skills and demographics. I’ve also committed to connecting my client teams to the tools they need to deliver data science and analytics together well. And there’s the rub. How does any leader know the best modern tools to offer top-notch talent?

I did some investigating at the Women in Data Science: Puget Sound event in April. The docket was full of interesting talks from some great women in data, and it was a “challenge” at every break to decide what to attend next! I especially loved the casual networking that happened between sessions. To spread the wealth for those of you who couldn’t attend but want to up your data leadership game, here are my top 4 takeaways. My fellow data leaders will likely appreciate these as you tap modern and mature platforms and build your own empowered teams.

Moving from scrappy to sophisticated with MLOps

One of the themes of the conference was maturing the process of how models are built and shared with others, particularly through Machine Learning Ops (MLOps). Several speakers highlighted how they had implemented principles of MLOps and how it enabled better teamwork and overall results. This perspective reflects many industry data science teams maturing from scrappy, siloed development into more efficient, collaborative work.

Making web apps as easy as web pages

Presenting outputs of a data science model comes in many forms; built into end products, encompassed in enterprise reporting tools like dashboards, among many others. One method, that has been a burden in the past for data scientists who weren’t software developers prior, has been hosting a Python-based web app. During this conference, we saw a demo of PyScript, an HTML-based method of deploying and running a Python app without a complicated server to support it. Now, if you can publish a webpage, you can publish a Python web app!

Debunking the unicorn myth

Another term I heard multiple times throughout the conference, from official talks and casual conversations alike, was around the perception of the “unicorn” data scientist: the person who can do everything. One of the keynote speakers presented a counterpoint to this type of data scientist, and focused instead on how data science is best done as a team. For individuals, this means building strong skills in areas that complement others on your team. For leaders, it means looking for and encouraging diverse areas of interest and growth among the data science team.

The future is shifting from building solutions to solving problems

This leads me to the thread through it all—design thinking. There was a call to shift from thinking about data science as a purely technical pursuit, and to instead take a more comprehensive approach focused on how an end-user’s problem will be solved. This reframing of data science keeps us focused on things that add value. As data scientists, we’re forever exploring, sharing, and learning ways to deliver data science outputs in an efficient and impactful way for industries, whether for employers or clients. With so many inspiring pathways, and the pool of diverse talent coming together, the future is bright. I can’twait to see how all of us, and especially data scientists who are women—bring it all to life. I’m counting on it.