Interview with Julian Troegel about the “Nala-AI” project

A river that wants to save itself – sounds like science fiction, right? The Nala-KI project is making this a reality: data, sensors, and artificial intelligence are being used to improve the water quality of German rivers. Our colleague Julian Troegel, software developer and data scientist at BREDEX, was part of the project and talks about how to combine currents, chemistry, and AI to create an exciting mix – and why patience is sometimes the most important data set.

Translated with DeepL.com (free version)

Julian, you are a software developer and data scientist at BREDEX—what specific role did you play in the project?

In this project, I was responsible for bundling and evaluating data and laying the foundation for future AI applications.

How would you describe the project in one sentence—in a way that someone outside our industry would immediately understand what was exciting about it?

In short: saving rivers with data. Nala AI uses sensor data to measure and analyze water quality and derive measures that restore balance to our rivers.

How would you describe the project in one sentence—in a way that someone outside our industry would immediately understand what was exciting about it?

In short: saving rivers with data. Nala AI uses sensor data to measure and analyze water quality and derive measures that restore balance to our rivers.

When did you realize that this project was going to be “special”?

It was clear from the very first project presentation that this was no standard IT project. I was immediately captivated by the vision of using data to prevent fish deaths and actively improve water quality. It was fascinating to see how many data points we could use and what could be achieved with them.

When you think back to the beginning, what were your expectations and which ones turned out completely differently?

To be honest, I thought we could start right away with existing data sets. In reality, it all started with the first sensor node that we installed ourselves in the river—and thus from scratch. But that’s what made it exciting: we literally saw the data flow. Especially with rivers, you need time to recognize patterns—rain, sun, seasons, or the fertilization of surrounding fields. Everything plays a role.

What was the customer’s goal and what was the sticking point where we could help?

The goal was to improve river quality. The customer was able to use a river near Cuxhaven for testing purposes through a subsidy program. The sticking point was combining various data sources—from weather data to tide calendars—in a meaningful way. We advised on which data might be relevant, what pitfalls there were, and how to build on this in the long term to develop AI that addresses real environmental problems.

What was the main problem or biggest challenge we had to solve for the customer?

The biggest challenge was to bundle and analyze the various data sources. Since real-time access was not possible, CSV files were used. In addition, data collection was still in its infancy.

What obstacles did you encounter along the way—and how did you (or the team) deal with them?

Since the sensor node was a prototype, there were regular failures because its development was being worked on in parallel. The sensors had to be calibrated, and we first had to find out which additional data sources were relevant to the project. In short, we experimented, adjusted, tested—and learned a lot in the process.

Was there a moment when you thought, “We can’t do this”? What did you learn from that?

No, I didn’t have that thought. The project was still in its early stages, and the most important task was to collect enough data and understand which of it was relevant. The bigger thought was, “We need more time” – because good data doesn’t grow overnight.

If you had to describe the project as a film genre, would it be more of a crime thriller, an adventure, or a comedy? Why?

Definitely an adventure. The quality of the river is the great unknown that we wanted to explore. It was about finding, understanding, and using data to make the environment a little bit better. An adventure with a laptop, a sensor, and a lot of patience.

If you had to describe the project as a film genre, would it be more of a crime thriller, an adventure, or a comedy? Why?

What noticeable improvements did the customer ultimately experience?

Various data sources were consolidated and aggregated so that they could be evaluated over a timeline. We also provided the customer with ideas and tips on what to look out for in the future—a kind of data compass for the next steps.

Is there any feedback that particularly sticks in your mind?

Yes, the customer expressed their heartfelt thanks for the voluntary work and regretted that we had left the project (for the time being). This shows that technological work leaves behind not only code, but also relationships and trust.

What was the most rewarding or proudest moment of the project for you personally?

Without a doubt, it was the thought of doing something positive for the climate, the ecosystem, and the rivers. Even if it was only a small contribution at first, it literally flowed into something bigger.

How would you summarize the success in one sentence?

Data sources have been identified and consolidated—and now the data collection for the next steps can begin. The first stone for the data foundation has been laid.

If you had to give the project a title, what would it be?

“First steps toward data-driven river quality.” Or in modern shorthand: Mission Nala – Save the Rivers.

What skills or tools were crucial for implementing the solution?

Python and Jupyter Notebook, especially with the pandas, requests, matplotlib, and sklearn modules. API connections to ecoweather, the German Weather Service, and the Federal Maritime and Hydrographic Agency were also used. Technology that understands nature – so to speak.

What specific improvements did the customer ultimately see?

The customer was able to build up expertise, promote exchange, and continue to use the program code we developed in Jupyter Notebook for connecting data sources. A real starting point for future AI projects.

What experience from this project will you take with you into future tasks?

How important it is to establish a stable data pipeline early on in new projects—because it is the backbone of any analysis. Without it, nothing flows.

What would you do differently today?

It’s hard to say—the project was too short to make any fundamental changes. Besides, it’s continuing in a different form, just without BREDEX. But who knows, maybe our data streams will cross again someday.

How would you describe the project in three words?

Innovative, visionary, climate protection

What new skill have you discovered in yourself?

I have learned to familiarize myself with a subject area that was completely new to me—river chemistry and water quality—and to develop ideas for practical applications based on this knowledge. It was like an expedition into a previously unknown data ecosystem.

If you could give other teams one tip for making similar projects successful, what would it be?

Familiarize yourself, understand, collect. Study the topic intensively, think outside the box, and collect as much data as possible. Only diversity reveals what is truly relevant.

How would you like to continue the project if you had a free hand?

Once more data has been collected, I would like to analyze it and develop initial algorithms to build an AI that can make predictions about river quality. An AI that not only calculates, but also thinks for itself—for the environment.

Conclusion

Data, sensors, and artificial intelligence—it sounds like modern technology, but there is much more to Nala-AI than that: it is a genuine environmental project with heart. Julian shows how data-driven innovations can lead to concrete solutions for our ecosystems. If you would like to learn more about how BREDEX develops sustainable and intelligent software solutions, please feel free to contact us.

Want more information? Then take a look at the website: https://nala-ai.org/