SEARCH TECHNOLOGY FOR COMPLEX QUERIES

One of the advantages of AI is that it simplifies specific searches. But what if you need to find something in a large database, while you don't know what you're looking for? Thoughtful Oasis' technology solves this. CEO Rens Dommerholt explains how their startup is ushering in a new generation of search.

The idea for Thoughtful Oasis didn't arise from a problem, but from the technology. After co-founder Kaspar Rothenfusser's first startup adventure at Cold Brew Crew, he became interested in ChatGPT. He followed his curiosity and delved into the workings of Large Language Models (LLMs). The idea arose to build a search engine for such complex models. He created a demo, presented it to Bekk Blando and Rens—whom he knew from Incubase—and from there, the three further developed the idea. "I knew Kaspar from his previous startup," says Rens. "When he presented the idea for Thoughtful Oasis to me, I was still working at Novel-T. Ultimately, I became so enthusiastic about the problems the technology could solve that I quit my job and now work full-time for Thoughtful Oasis."

Efficient information disclosure

The problems Rens is referring to are very diverse. "If you look closely, you'll see that this search problem occurs in every industry," he says. The team found the most striking example at the fire department, where they were conducting a pilot. "Many firefighters are currently retiring, and fewer young firefighters are joining. So they have to do the same work with less capacity—and money. It's therefore important that they work as efficiently as possible." In practice, it works like this: A commander receives a notification via the pager. They then cycle to the control center, where the suit is already waiting. Once the commanding officer gets in the vehicle and leaves the station, the team has about two to six minutes to assess the situation on site before they arrive. And in those minutes, Thoughtful Oasis can make all the difference.

Rens: “During the drive, the team views the messages from the control room on a tablet (MDT), which they also have a direct connection to via their walkie-talkie. Right now, that's all they have to assess the situation—besides their training and experience. What our system adds value to is its extensive database of training documentation, previous simulations, protocols, land registry information, and more. Our tool takes human factors into account so firefighters aren't overloaded with information. They receive the most important information, ensuring nothing is overlooked, such as asbestos or the number of casualties. This allows them to make better decisions faster, increasing safety.”

Various complex searches

This is one example of how technology can make all the difference. But Rens is also looking at other application areas they want to explore. “I'm thinking of the pharmaceutical industry, where our search method helps with cluster analyses. And yet another very different application can be found at universities seeking to secure European grants. Among the thousands of people working at a university, it's sometimes difficult to find the right researchers for various projects. Our tool can find those people and identify connections between other researchers, which can lead to new opportunities. That's another complex search we simplify.”

Experience and Opportunities

The team has a good amount of experience. Kaspar completed the START and ADVANCED Programs with his previous startup, and Rens supported startups through Novel-T for five years. “Still, we've benefited greatly from the support of Novel-T,” says Rens. “For example, we had the opportunity to give a demo to AI software developers at AI Tinkerers during Enschede Slush’D. We were also able to participate in pitch competitions and international missions. Furthermore, we received the TOP loan, and the network I built through Novel-T is a huge help. Several projects are even coming out of that now.”

Climate-conscious

The young company thus has a unique piece of technology at its disposal that is widely applicable. However, AI isn't always the holy grail due to the energy consumption per search. A ChatGPT query uses about 10 times more energy than a Google search. How does that work at Thoughtful Oasis? "When you make a dataset searchable, you always have to perform a processing step that consumes energy. We estimate that our processing is more extensive and therefore more demanding. We hope to compensate for that by searching more effectively and efficiently, resulting in fewer searches. We still need to test that assumption, but we're very aware of the burden on the climate.”

Big ambitions

Within the next two years, the team wants to identify the most pressing problem for their solution. “From then on, I hope to scale significantly,” says Rens. “2035 is still a long way off, but we hope to be a major player in Europe and America by then. Who knows, we might even have a few spin-outs here and there that will further develop our search technology for specific markets. We're quite ambitious, and rightly so, given our technology. For example, EIT Digital calls us the stand-out case in the field of AI technology, and we won the Hackathon Open Government ‘Terminal WOO’. Of course, we still have to prove ourselves in the coming years, but I'm convinced we'll succeed.”