In the Author Spotlight series, TDS Editors talk to members of our community about their work in data science and AI, their writing, and their sources of inspiration. Today, we are excited to share our interview Sara Nobrega.
Sara Nobrega is an AI engineer with a background in Physics and Astrophysics. He writes about LLMs, time series, workflow, and AI-enabled workflows.
He holds a Master's degree in Physics and Astrophysics. How does your background play into your career in data science and AI engineering?
Physics taught me two things that I always rely on: how to stay calm when I don't know what's going on, and how to break a scary problem down into smaller pieces until it's less scary. And… physics really humbles you. You quickly learn that “intelligence” means nothing if you can't explain your thinking or reproduce your results. That thought is probably the most useful thing I can bring to data science and engineering.
You just wrote a deep diving in your transition from data scientist to AI engineer. In your day-to-day work at GLS, what is the main difference in mindset between the two roles?
For me, the biggest change was going from “Is this model good?” to “Can this system survive in real life?” Being an AI Engineer is not so much about the perfect answer but more about building something reliable. And honestly, that change was uncomfortable at first… but it made my job feel so useful.
You noted that while a data scientist may spend weeks developing a model, an AI engineer may only have three days to implement it. How do you balance excellence with speed?
If we have three days, I'm not in a hurry to make small improvements. I am chasing confidence and honesty. So I'm going to focus on a solid foundation that already works and an easy way to monitor what happens after launch.
I also like to be sent in small steps. Instead of thinking “use the last thing,” I think “use the smallest version that creates value without creating chaos.”
How do you think we can use LLMs to bridge the gap between data scientists and DevOps? Can you share an example where this worked well for you?
Data scientists talk about testing and results while DevOps people talk about reliability and repeatability. I think LLMs can help as a translator in a practical way. For example, making tests and documentation so that what works on my machine is “working in production.”
A simple example from my work: if I'm building something like an API endpoint or processing pipeline, I'll use LLM to help me write the boring but important parts, like test cases, edge events, and clear error messages. This speeds up the process considerably and keeps the motivation going. I think the key is to treat the LLM as a little kid who is quick, helpful, and occasionally wrong, so reviewing everything is important.
You did it research cited suggests a significant growth in AI roles by 2027. If a young data scientist could learn one engineering skill this year to stay competitive, what would it be?
If I had to pick one, it would be learning how to post your work more often! Take one project and do something that can work reliably without you supervising it. Because in the real world, the best model is useless if no one can use it. And the people who stand out are the ones who can take an idea from a book to a real thing.
Your most recent work has focused on LLMs and time series. Looking ahead to 2026, what emerging AI topic are you most excited to write about next?
I lean more towards writing about the AI workflow (how you go from an idea to something reliable). Besides, when I write about a “hot” topic, I want it to be helpful, not just entertaining. I want to write about what's working, what's happening… The world of data science and AI is full of trade-offs and ambiguities, and that's taken me by storm.
I'm also curious about AI as a system: how the different pieces fit together… stay tuned for this year's articles!
To learn more about Sara's work and stay up to date with her latest articles, you can follow her on TDS or LinkedIn.
