How this data scientist keeps up to date in a rapidly changing industry

Data science is constantly evolving, but Liberty IT’s Naomi Hanlon explains how she’s keeping pace and shares her tips for early-career data scientists.

Having worked as a data scientist for over four years, Naomi Hanlon has experience in various industries, from manufacturing and mobility to insurance. She currently works at Liberty IT.

For her, a typical day starts at 8am, when she checks her emails and starts working on the tasks for the day, as well as a short meeting with her team.

“Depending on the type and stage of the project, the rest of my day will include a combination of meetings, data cleaning, exploratory data analysis, feature building, presentation building, presenting, collaborating, writing reports, code reviews, code walkthroughs, and integration.”

Hanlon is currently trying hybrid work, which means she spends half of her time at home and the other half in the office. She also works a compressed week, meaning four longer days and Fridays off.

“Data science is a rapidly evolving field, so trying to keep up with the latest research is an impossible task”
– NAOMI HANLON

What types of data science projects do you work on?

One of the first projects I worked on at Liberty IT was migrating existing models from SAS to Python. This doesn’t seem like the most exciting project from a data science perspective, but it was one of my favorites because I learned a lot from this project and it helped me in subsequent ones .

Being paired with an engineer allowed both of us to learn a lot about each other’s field. As a result, I was writing much cleaner code – modular and with (shocking) unit testing.

We had a consistent branching strategy in Git and performed code walkthroughs, pull requests, code reviews. Collaboration between data scientists and engineers is the norm for us now, but it was all new to me at the time. This project changed the way I write data science code.

Most recently, we worked on our company’s first continuous learning pipeline. Typically, model performance declines over time, due to changes in customer appetites, needs, or tendencies.

Click here to discover the top science and technology employers hiring right now.

Usually, these less performing models will require data scientists to retrain them using more recent data. It takes time and requires a data science resource. Continuous learning solves this problem.

It is essentially a repeatable model that allows models to adapt in production. It uses the most recent data to frequently recycle the model. This improves efficiency because data scientists can use their time to answer new business questions rather than readjusting models. It also ensures that the models stay up to date and provide more reliable forecasts.

We had to use new (to us) tools like Managed MLflow and Luigi to build the pipeline and the result is a repeatable model that can be extended to other areas.

What skills do you use on a daily basis?

Technically, I would rely on Python most of the time. Although I love R, Python seems more accessible for cross-functional teams.

Communication is often seen as an important skill in many different roles. For data science, this means having to effectively communicate your findings or approach. The unexpected skill is being able to present your information at the right level for your audience.

Some stakeholders will want the high level findings and conclusions, an executive summary. More technical audiences will appreciate the granular information – what metrics, packages and methodology were used. Engineers will often enjoy going straight into code.

Business needs drive data science. The ability to develop a solid understanding of the problem, translate the question into an experimental design, and come up with potential solutions that are part of the larger business strategy is essential. This allows us to keep a job as data scientists.

What are the hardest parts of working in data science?

Data science is a rapidly evolving field, so trying to keep up to date with the latest research is an impossible task. Blogs and podcasts are a great starting point and make new concepts accessible.

At work, we have a number of knowledge sharing sessions within our data science teams to allow people to share anything interesting. It can be something they encountered or worked on. It has also been helpful in developing best practices as we continually share work and feedback.

I have already mentioned the importance of communication in data science. One aspect of communication that I find difficult is presentation. Working remotely has been the best thing for me where I can have on-screen prompts as support during a presentation.

Over the years, many people have told me that the more you do it, the easier it gets. I hated that. I also don’t like the fact that they were probably right. In my current role, I’ve presented more than ever and it really does get easier with exposure. Working remotely really can’t be underestimated, it’s awesome.

Do you have any productivity tips that help you throughout the day?

I have a very bad memory, so I rely on notes. Using a few minutes at the end of the day to make a list of upcoming tasks helped.

Like a real data geek, I also started tracking my habits every day. These are things I want to spend my free time doing like exercising, reading, and listening to podcasts. Phones can be so time consuming, where an hour can pass with nothing to feel good about afterwards. Habit tracking, so far, has helped me spend my time doing something a little more interesting.

Having these goals as a habit also helps productivity at work. I find that I’m much more likely to go for a walk or read a chapter about my lunch, just for the simple fact that I can follow it. It gives me a real break from screens and I can add another data point to habit tracking. Win-win.

What skills and tools do you use to communicate with your colleagues on a daily basis?

We weren’t immune to the Slack v Teams debate, but we landed on Teams for project work. It’s more accessible enterprise-wide.

I’ve talked about knowledge sharing sessions and code walkthroughs – they all take place on Teams with screen sharing. We use Slack, but it’s generally more informal communication with channels for cooking, exercising, gaming.

What do you enjoy most about working in data science?

Deep down, I really like trying to work things out. This includes puzzles, riddles, Rubik’s Cubes – anything. It’s also probably why I love crime dramas and thrillers, the investigative element and participating in unraveling what’s really going on. That’s a big part of what I can do as a data scientist – there’s a question and we’re using data to try to answer it and solve the problem.

Specifically in this role at Liberty IT, I have a lot of autonomy over my work. This gives space to experiment, learn and iterate, which is important in data science.

Generally, we work on short-term assignments, which means that my work is also varied. I manage to apply different approaches and techniques depending on the field of activity and its objectives.

What advice would you give to someone who wants to work in data science?

My first piece of advice is to get involved in data science. Wave, I know. We all come to data science with different backgrounds, so find something that matches your level of interest and the time you have to devote to something new.

As an entry point, there are many online resources you can tap into. Tutorials, courses and YouTube videos. With that, there are many open datasets available to help you get started with the basics, reading data, visualizations, exploratory data analysis.

Beyond the basics, one of the best things you can do is start trying to use a real data set to answer a question that interests you. Tutorials and toy datasets will get you this far, but start building those skills with a project you set yourself.

Being part of a larger community will keep you engaged in the world of data science and is a great way to learn more. There are many ways to get involved, including online communities, meetups, forums, and Kaggle.

Making connections and learning from the day-to-day work will help guide you towards more specific goals. This will help you determine the skills and qualifications you need as you progress through your data science journey.

Don’t miss out on the knowledge you need to succeed. Sign up for the brief dailySilicon Republic’s must-have science and technology news digest.

Sean N. Ayres