Interview with Irene Ternes, Co-Founder and CEO at DataMonkey

 

In this FFplus Interview, we spoke with Irene Ternes, Co-Founder and CEO of DataMonkey, a German tech company dedicated to making AI-powered data more accessible for business experts, data analysts, and IT teams.

We discussed the most transformative trends currently reshaping data analysis and visualization—particularly within the geospatial domain—and explored how DataMonkey is pioneering the use of GeoAI, combining large language models with geospatial intelligence to make map-based data more intuitive and insightful. Irene also shared insights into training LLMs for geospatial data and what this means for the future of data-driven decision-making.

Irene Ternes, Co-founder and CEO, DataMonkey.

Copyrights: DataMonkey

 

To start, could you briefly introduce DataMonkey?

Irene:

DataMonkey is a geospatial data company based in Berlin, Germany, meaning we work with any kind of data that can be represented on a map. Essentially, we use AI – particularly large language models – to make geospatial data accessible to business experts, enabling users to find, retrieve, filter, and combine geospatial datasets using natural language.

 

From your perspective, what are the most transformative trends that are currently reshaping data analysis and visualization? Particularly within the geospatial domain?  And how are these trends influencing decision making, innovation and interaction in this field?

Irene:

I think there’s a lot happening in the world of data right now. In particular, the rise of natural language models over the past few years has had a tremendous impact on how we can make data more easily accessible – especially for people who don’t come from a data background but work more on the business side.

Essentially, using natural language helps bridge the gap between data and humans, making information easier to understand and access. We’re seeing this trend in the geospatial field as well – when working with map-based data, combining it with natural language is becoming a crucial element.

Ultimately, the key challenge is to make data truly accessible for end users by building solutions that meet very specific business needs and workflows. It’s about closing the gap between what’s technically possible and what’s genuinely useful for companies and teams, in order to create real value.

 

DataMonkey is pioneering the use of GeoAI by combining cloud language models with geospatial intelligence to make map-based data more accessible. Could you walk us through how AI is integrated into your workflows and solutions? And can you introduce examples of how your clients are using their capabilities to solve real-world problems and make better decisions?

Irene:

I would say that one thing which makes geospatial data quite special compared to other types of data is that users often need to work with both internal and external datasets. For example, if you think about energy providers or energy development companies, they frequently rely on external data provided by local or regional governments – such as information about where wind turbines can legally be built or where it would make sense to install solar panels.

We see this very often: teams need to combine their internal data with trusted external sources. These can come from government bodies, open platforms such as OpenStreetMap, NGOs, or other reputable organisations. Bringing these different sources together is what makes geospatial data particularly unique.

Where we see a lot of potential is in improving data retrieval – helping users find the right datasets, identify trusted sources, and tailor or combine data to address specific questions. This makes the process much easier, especially for people who are not familiar with GIS (Geospatial Information Systems). By doing so, we can help users save a great deal of time and lower the barriers to accessing and using geospatial data.

That also brings us to who we actually work with. Most of our direct touchpoints are with geospatial experts – for instance, GIS analysts or scientists. However, this often leads us to collaborate with business departments, which are the ones needing to answer specific questions or streamline particular workflows.

In the end, it always depends on who the end user is and how technically skilled they are. But our ultimate goal remains the same: to make their work easier. That’s exactly why we’ve developed what we call our Geo LLM – to simplify and enhance how people interact with geospatial data.

 

So, let's now talk about your experience with participation in the FFplus sub-project. Could you briefly introduce your use case and what challenges it addresses? And especially what innovative solution have you been developing?

Irene:

We are very much focused on training large language models (LLMs) for geospatial data. The reason behind this is that geospatial data has two distinct components. On one hand, there’s the language element – for example, streets, cities, or specific objects all have names, which appear in natural language. On the other hand, there’s the numerical component, such as geographical coordinates and other numeric values.

Because of this dual nature, a standard, out-of-the-box LLM is usually not very effective at answering geospatial questions or providing reliable, high-quality data outputs.

Within the FFplus programme, our work focuses on training LLMs to better understand geospatial data. We do this by using anonymised user data from our platform, as well as synthetic data, to build a kind of translation layer between the numerical values on the geospatial side and natural language.

This allows users to simply type a query on our platform — for example, “Show me all solar plants in Berlin” or “Show me all bike lanes in Amsterdam” — and the AI will identify the correct dataset from a trusted source, filter it for the requested region, and deliver the results in a reliable and trustworthy format.

 

What key business outcomes do you expect the solution to deliver once the sub-project is completed? So how do you see it impacting your operations, corporate competitiveness and market opportunities?

Irene:

I would say that for us, there are two main aspects. On the one hand, this gives us a unique opportunity to improve our AI model, because in the end, data quality is absolutely essential – not only for many use cases in general but especially for geospatial data. If users cannot trust the data, it simply becomes useless. So the first part is about ensuring that the results are truly reliable and that we can deliver trustworthy datasets to users. This will, of course, also enable us to apply our solution to a wider range of projects, clients, and use cases.

The second aspect, which I find particularly interesting, is that in a later phase we’ll also be able to layer different datasets on top of this foundation. Having this translation layer between geospatial data and natural language makes it possible, for example, to integrate satellite data, which can be extremely valuable for other use cases depending on the client’s needs.

 

Your sub-project has been running for months now. So as the sole partner working in this case, have you encountered any specific challenges and how you've been approaching solving them?

Irene:

Exactly. So we are the sole partner, but we receive excellent support from the local HPC team, which we are very grateful for.

We have definitely encountered some challenges throughout the programme. For example, one thing we hadn’t anticipated at the start was that EUROHPC is somewhat of a closed system. It doesn’t, for instance, provide internet connectivity. We realised this only later, but we soon understood that at some point, we would need internet access to validate the data.

We were able to train the AI model on the supercomputer, but eventually, we needed to perform evaluations to ensure that the parameters were correct and the results reliable. Without internet access in our setup, this wasn’t possible, so we had to find a workaround to make it work.

The EUROHPC setup is obviously very powerful, but also quite unique. I think it’s important to consider such processes and setups from the start to ensure everything functions smoothly.

Overall, though, it’s been a fantastic opportunity for us.

 

Let's wrap up this interview with some final thoughts. What advice would you give to the SMEs that haven't explored the use of HPC and AI in their business?

Irene:

I would say it makes sense to look at very specific use cases and carefully verify what you actually need. Depending on the case, you might be able to apply an out-of-the-box AI model or large language model, or you may require a pre-trained model or even a more complex AI system or agent setup.

It’s important to examine your current position, understand your actual requirements, and determine where it truly makes sense to train a model. By doing so, you can leverage the resources available in the European Union to create tailored knowledge and train a model with the specific data you want to use.

When applied this way, the results can be very powerful and make a real difference, compared to simply using generic, out-of-the-box solutions.

 

And this brings us to the end of this interview. Irene, thank you very much for your valuable insights. We wish you all the best and enjoy your future endeavours. 

 

Learn more about the sub-project “Geo-Llama, the first LLM for Geographic Data” here.