The world’s leading publication for data science, AI, and ML professionals.

If You Need to Get Unstuck, Try a Different Angle

Not sure how to move beyond a thorny data science problem? Break it down to smaller pieces-and ask for a peer's perspective.

Author Spotlight

In the Author Spotlight series, TDS Editors chat with members of our community about their career path in data science, their writing, and their sources of inspiration. Today, we’re thrilled to present our conversation with Carolina Bento.

Photo courtesy of Carolina Bento
Photo courtesy of Carolina Bento

Carolina is a data scientist with a special interest in Data Visualization. Before dropping out of a PhD program where she focused on data analysis and data visualization in large-scale networks, she completed a Master’s degree in Computer Science. She’s been working on data science and analytics ever since.

When she’s not writing about the fundamentals of data science and data visualization, you can find her reading about science, history and behavioral psychology, doing calisthenics exercises, or playing lo-fi electronic music.


Thanks for joining us, Carolina! Let’s start with some basics: how did you find your footing within the wide world of data science?

When I started the computer science Master’s program, I needed to narrow down the scope and focus to a couple of areas for the major and minor.

Even though data science wasn’t well defined at the time, I naturally gravitated towards courses that were all about cleaning, analyzing, and making sense of data.

Later on, during the course of my PhD, the excitement and curiosity for data science really kicked in. The field was starting to get traction, I was following the research closely, and got the chance to dive deep into data analysis and data visualization.

I eventually dropped out of the PhD, but this experience was so important that I’ve been working on data science ever since.

When you encounter a challenge in your data science work, how do you go about tackling it?

Whenever I get stuck while learning something new or working through data science problems, I always go back to asking the same set of questions.

  • Did I break this problem into atomic pieces? Sometimes I may rush into working on a problem that is too big and with too broad a scope. In these situations, I nudge myself to take a step back and check if I can break the problem into many smaller, individually addressable tasks. It reduces the scope of the problem I’m working on to something narrowly defined that can be tackled with more specific algorithms and techniques. It may not necessarily reduce the complexity right away, but it helps me stay focused.
  • Can I approach this problem from a different angle? Data science, like any other field, benefits from a good dose of creativity and from trying things that might not be obvious at first. Sometimes, when I get stuck, I think about how I can approach the problem from a different angle. What other questions can I ask about the specific problem that may lead me to think about it in a different way, and, ultimately, to a solution I haven’t thought about before?
  • Have other people already worked on something similar? This usually means talking to my peers or experts in the area, doing additional rounds of research to see if someone has stumbled on the same or similar issues before, and check if there’s additional information about the specific techniques I’m trying to use.
  • What am I not seeing about this problem? When I get stuck I try to understand if I have all the context necessary, and if not, I review my hypotheses and assumptions. Sometimes I realize my initial hypothesis was not the best, so I refine it and start over. But I’ve found that talking to my peers, explaining the problem I’m working through, also helps me get unstuck. It makes me think thoroughly about the problem in order to articulate it. More often than not, I come out of those conversations energized and with new ideas—either because my peers had great suggestions, or simply because the process of explaining it to someone who didn’t have all the context made me think more creatively about it.

This is my way of getting unstuck, but I know it’s far from being a perfect solution. With every new problem I find myself looking for ways to improve upon it, which is also an interesting learning process.

What inspired you to write about data science concepts to a broader audience?

Data science is a very hot field right now. A lot of new research, frameworks, and applications come to light every day, so it’s easy to feel out of the loop.

I wanted a way to revisit core data science concepts and continue to expand my knowledge. But when I started revisiting materials and trying to learn new things I realized that, even in introductory resources, there was a good amount of abstraction that made it hard to follow.

Some people are very good at understanding abstract concepts right away, but I find it much easier to learn something when I have one or two detailed examples that hold my hand through the entire process. Then I can step back and start thinking about that abstract concept.

I decided to write articles because it was the medium I was more comfortable with. I used to spend the summers writing short stories when I was a kid, so writing felt more natural.

But writing also fits my preferred learning process. It forces me to think deeply in order to express myself in a way that is easy to understand. It’s a skill I’ve always wanted to improve on.

On top of being a good way to continue learning, it is also an opportunity to contribute to the pool of in-depth data science resources.

You often add beautiful custom illustrations to your posts. What do you think they bring to the learning process?

Creating the illustrations is super fun to do!

They’re my way of giving a peek into the main theme of the article, and to introduce a few visual check-ins throughout the piece.

At the core of the article is the story. The story helps me break down the topic into its core concepts, and guide the reader through all the details. Charts and diagrams are other visual learning elements I like to use, so the reader can gradually build their mental model as they’re following the story.

Images guide me during the learning process too. I take a lot of notes when I’m learning something, but I also scribble a lot. I found that when I can see what I’m learning, I can establish a stronger intuition and deeper understanding about the topic, and can also remember it a bit better.

I try to write the articles as if they were my curated and organized collage of notes and scribbles, so anyone can follow along.

What are your hopes for the data science community in the near future? What kind of change would you like to see?

Data science and machine learning are such broad fields, with so much implicitly required knowledge, that I’d love to see more introductory resources that guide someone who’s just getting started.

Not to mention that starting fresh in a new field is daunting!

There has been a lot of progress on this lately with different articles, videos, and courses, but everyone learns a bit differently.

So I’m looking forward to seeing how we can make data science and machine learning more accessible and approachable to anyone who’s starting out.


Curious to learn more about Carolina’s work and data science interests? You’ll find her entire archive of posts on her Medium profile, and you can follow her on Twitter for updates and recommendations. Carolina has published a wide range of excellent tutorials and explainers on TDS over the last few years; here is just a small selection of highlights.

  • Central Limit Theorem: A Real-Life Application (TDS, October 2020) A typical post from Carolina breaks down a complex concept in statistics, data science, or machine learning through a relatable story that most of us can visualize with minimal effort. Read this introduction to the central limit theorem to see how she does just that with the example of a grocery store manager who needs to decide how much seltzer to order to keep stocks at the right level.

  • Logistic Regression in Real Life: Building a Daily Productivity Classification Model (TDS, March 2021) Many of us might occasionally confuse linear and logistic regression. That will be far less likely after digesting Carolina’s clear walkthrough, which uses daily routines and personal productivity as a vehicle for explaining a tricky topic.

  • Markov Models and Markov Chains Explained in Real Life: Probabilistic Workout Routine (TDS, December 2020) If you’re taking your first steps in data science, you might find the abundance of new names and terms overwhelming. Here, Carolina tackles one of the most important ones, bringing clarity and accessibility to Markov models and chains.


Stay tuned for our next featured author, coming soon. If you have suggestions for people you’d like to see in this space, drop us a note in the comments!


Related Articles