Data science engineer: A day in the life

Jeffrey Partyka, principal data science engineer at Xactly, shares insights and debunks myths on the daily reality of this dynamic role
1 reader likes this.

A common narrative for a day in the life of a data scientist is that we’re building the next cutting-edge machine learning (ML) model, showcasing it at conferences, and soaking in the applause. However, this is far from the daily reality for most data scientists.

Data science is much more pragmatic in reality: Instead of building new models from scratch, we fine-tune well-known models from well-known libraries and choose predictors based on domain knowledge. From this, we can provide value quickly.

To debunk the myths, here is what my career as Principal Data Science Engineer at revenue intelligence company Xactly is like:

My daily responsibilities

The daily responsibilities of any data scientist often depend on team size. On bigger teams, roles tend to be more specialized. There will be multiple people who work on a particular phase of the data science lifecycle. On a smaller team like mine, you must wear multiple hats and understand all phases.

[ Also read Data scientist: A day in the life. ]

My day starts with checking in with my teammates on the different models they’re working on or problems they’re solving. On occasion, all is well, and we celebrate that. But customarily, there are numerous challenges to address.

After connecting with my team, I work on my models and projects, which tend to be tougher and more complex. I try to dedicate the majority of my time to solving challenges ranging from bolstering model performance to solving deployment issues.

Show the value of data

Between communication, data engineering, meaningful result reporting, and more, data scientists have many goals. At Xactly, my daily goal is to illustrate to the rest of the organization and our customers the value of our data.

Strategy and evangelization are a huge priority. It’s important to illustrate how data science is useful in other departments like engineering, marketing, customer experience, and sales. In the space of a day, this can be messy, requiring us to dig into the details of how data was created. From this, we hope to create new predictors that could be incorporated into our models.

My team focuses on solving various technical problems across the organization daily. Over time, each day’s work contributes to achieving bigger goals. I see it as solving one or two subproblems per day, which over time, feeds into solving a larger problem that serves a bigger purpose.

As we finish projects, we build on that success by developing new models and making new insights. For example, a recently deployed model achieved sales forecasting accuracy of nearly 100 percent. Now we’re building that same capability into our other models.

How I embrace hybrid work

Xactly appears to be encouraging a hybrid work model, so when I have days that focus on collaboration and communication, I find it is more effective to work in person with my team and stakeholders.

[ Related read Remote work: How to balance flexibility and productivity ]

On immersive technical days where I need zero distractions, I’ll typically work remotely. Working remotely is more productive for diving deep into the data and solving complex problems. Mathematics, analysis, and software engineering are all core components of my work-from-home days, and the fewer distractions, the better.

Data science is continually evolving

For data scientists, the challenge isn’t necessarily how to build models or write code; the key question is what models to build and what code to write.

Since I began delving into data science back in the 2010s, a lot has changed. Notably, expectations are higher – and we have automation available to meet those expectations.

Previously, the big expectation for data scientists was to build a model. That alone was often enough to satisfy stakeholders, never mind explainability. There wasn’t a driving need to answer questions about the reasoning of the predictions because the idea of building an ML model that spits out predictions was downright futuristic and held enough promise on its own.

However, over time, this led to questions about the correctness of the predictions This resulted in an influx of explainability frameworks showcasing the influence of model predictors. This is where that evangelist responsibility starts to come into play.

Eventually, increased automation and community support arrived, paving the way for data science to be propelled to new heights. But this leads to another misconception, which is that there eventually won’t be a need for data scientists because automation will replace us.

I disagree with this.

For data scientists, the challenge isn’t necessarily how to build models or write code; the key question is what models to build and what code to write. The challenge is figuring out the best model to build that brings the most business value. And to figure this out, you need company knowledge that AI doesn’t have.

It’s predicted that data science will see more growth than almost any other field between now and 2029. The data science profession is not going away anytime soon, so if you’re looking to pursue a career in this industry, I’d suggest you block out the misconceptions surrounding it. My advice is to find a project idea that’s of great interest to you and build it from scratch. Mix this hands-on experience with some education via online courses and books, and you’re on your way to a successful data science career.

[What is a day in the life like in your role? If you’d like to participate in this series, reach out here.]

Jeffrey Partyka has been a principal data science engineer at Xactly since July of 2020, actively conducting and overseeing numerous machine learning/statistical projects relating to sales forecasting, churn, cross-selling and other areas of sales.