Adam Novozámský is the supervisor of the internship Segmentation of Biological 2D Objects, which you can apply for at the Institute of Information Theory and Automation of the CAS (ÚTIA). This is where mathematics, computer science, and engineering come together to develop smart solutions to complex real-world problems in medicine, transportation, and industry. ÚTIA has a long-standing focus on topics such as artificial intelligence, computer vision, and statistical modeling. What can this internship offer you? How can AI be applied in everyday life and how can it support scientific research? Read on to find out.
Is there a common theme or approach that connects the internships offered at your institute?
Most of our internships revolve around working with data – but not in the “just upload it into a model and hope for a miracle” kind of way. It’s about deep understanding, creating your own models, experimenting with algorithms, and thinking critically about what the data actually means. Every project has a practical dimension, but it also leads to fundamental questions like: How do we know when we truly understand something?
What knowledge and skills should students have if they want to apply for internships at your institute, or specifically for your internship?
To take part in internships at ÚTIA, you need some foundational knowledge, such as programming – ideally in Python, since it’s most commonly used for working with data, machine learning, and visualization. We don’t expect students to know every library by heart, but they should be able to think algorithmically and not be afraid to debug their code.
A basic background in mathematics is also very useful, especially linear algebra, statistics, and probability, as these are key tools for understanding how and why algorithms work the way they do.
We also offer internships focused on working with biological data, such as microscopic images, so openness to interdisciplinary collaboration is definitely welcome.
In your opinion, in which areas can AI most significantly simplify scientific work, and why?
AI today is more than just a smart tool for filling out spreadsheets or recognizing images. It can support scientific work in three key areas:
Routine tasks
Processing experimental data, analyzing images, searching for literature, all of these are tasks AI can handle faster than a human. That doesn’t mean scientists are out of a job –rather, they can focus more on meaningful questions and spend less time clicking around.
Decision support
Researchers are often overwhelmed by vast amounts of complex data, and AI can help identify patterns or potential solutions that aren’t immediately obvious. This is especially valuable in fields with a lot of noise or uncertainty – such as biology or social sciences.
Idea generation
Surprisingly, AI can even be a source of inspiration – for instance, by proposing alternative hypotheses or novel ways of testing something. It doesn’t replace human creativity but can give it a helpful push.
And let’s not forget the less glamorous but very exhausting part of research: administration. Reports, annotations, summaries… AI can ease this burden significantly and save a lot of time.
Do you see any major limitations or risks in using artificial intelligence in scientific research? For example when handling sensitive data or interpreting results?
Absolutely. AI in science isn’t just a magic helper, it has its limitations and risks that are important to be aware of.
The first issue is the “black box” effect. If I don’t understand why a model is recommending something or how it reached a certain conclusion, it’s hard to build a credible scientific argument on that basis. In research, it’s not enough to say “AI said so” – we need to understand why.
The second big challenge involves sensitive data – in medicine, biology, or social sciences. This isn’t just a technical issue but also an ethical one, involving participant consent, privacy protection, and legal frameworks.
And the third risk is overestimating AI. Models can find patterns, but they don’t understand what those patterns mean in the real world. That can lead to misinterpretations.
AI is a powerful tool, but it’s still just a tool. Critical thinking and scientific curiosity must remain the priority.
Your internship focuses on the analysis of biological data in collaboration with microbiologists. How does this collaboration work in practice?
Our collaboration with microbiologists is not a one-off consultation – it’s more like a long-term dialogue between two very different worlds. Microbiologists provide us with microscopic data, often in the form of 2D or 3D images, but even more importantly, they provide context: what we’re actually seeing in the image, what is relevant, and what is just noise.
Our task is to develop algorithms that can detect meaningful structures in this data – for example, identifying cellular structures, classifying types of microorganisms, or tracking their development over time. This typically involves training models on labeled data and then testing them on new samples.
What challenges have you encountered in combining computer science and biology?
The biggest challenge is that computer scientists and biologists often speak very different languages. A computer scientist focuses on model accuracy, while a biologist is concerned with clinical relevance. The computer scientist tunes hyperparameters, while the biologist thinks about cell cultures. Sometimes it’s difficult to explain to each other what “good enough” means in each field. So the key is patience, mutual curiosity, and a willingness to learn from one another.
What specific problem in microbiology does your project address?
One of the specific projects we’re working on is the automatic segmentation of spheroids, which are 3D cellular structures used, for example, in cancer drug development. The goal is for the algorithm to automatically detect the shape and size of a spheroid in microscopic data, enabling researchers to monitor how it reacts to different substances.
This is a great example of how AI can support research with real-world impact in medicine.
Can students apply the knowledge gained from this internship to everyday life? In what situations might knowledge of computer vision and AI be useful?
Definitely and probably more than it might seem at first. The knowledge students gain isn’t just about programming or training models. It’s mainly about learning to think algorithmically, work with data, and formulate problems in a way that can be solved intelligently – a skill that’s transferable almost anywhere. That’s often the most valuable thing students take away from our internships.
As for computer vision, the same principles are used in quality control in manufacturing, medical imaging analysis, or security systems and autonomous driving. AI can also help in other professions by automating routine tasks, from filtering emails to generating materials for work or school.