a day in the life at the eth hackathon

by Oscar Sodian (Visiting Analyst at redalpine)

I recently joined the Swiss AI Hackathon hosted by the ETH Entrepreneur Club. Bright minds from ETH, EPFL, and all across Europe gathered in Zürich for 24 hours to turn raw ideas into something tangible. And for me, it was a chance not just to build a product in 24 hours and compete, but to show how we thrive under pressure at redalpine and connect with the next generation of outstanding European founders.

I teamed up with Pavel (another Visiting Analyst at redalpine), a computer science student from ETH, and a fusion engineer from Italy. A diverse mix from different fields, but that’s exactly what we needed in order to be exceptional. We started the day with big ambitions. First, we brainstormed about ideas and quickly realized we needed to build something extremely unique in order to be able to compete. Our strategy was clear from the beginning though; tap into a niche market, dominate and expand from there. 

After breakfast, many coffees, and a desk lunch, we finally mapped out our idea. We wanted to build something real, something that can actually be used later on. And that’s why we built Lepton, an AI-powered construction planning platform that optimizes workflows, reallocates workforce dynamically and transforms technical drawings into actionable, risk-aware work plans. We implemented an LLM capable of extracting potential tasks directly from project documentation. Based on this output, an AI system then categorized these tasks into predefined groups and ranked them by priority. Subsequently, the algorithm generated a comprehensive task map for the construction site and assigned tasks to workers according to their skill profiles.

Naturally, construction sites are prone to unexpected disruptions. When such issues arise, workers can report updates in real time. The AI processes these updates, identifies the underlying problem, recalibrates the task list, and the algorithm adjusts the overall project plan accordingly. Processes that traditionally take three to four weeks were reduced to a matter of seconds with Lepton.

Our algorithm works as follows: We calculate a directed acyclic graph (DAG) using topological sort. We considered all tasks that have no dependencies. Now, the problem was to assign workers different roles to their possible tasks. The underlying problem corresponds to a maxflow problem, which we solved using Edmonds-Karp polynomial algorithm. After a fixed time, the completed tasks are deleted from the dependencies graph and new independent tasks are added to the network.  Moreover, we implemented some tweaks using binary search for a speed up of some internal processes. The aforementioned subroutine runs in polynomial time, henceforth, is scalable for a large number of tasks and workers.

While Lovable developed the front end and the AI integration, achieving seamless end-to-end performance required deeper backend work in Python, particularly for the algorithmic logic and system orchestration. For the cherry on top, we even generated live risk heatmaps from CAD files and reprioritized the tasks accordingly with the algorithm. 

Of course, the journey wasn't smooth sailing. By 9pm, almost 13 hours into the hackathon, we had faced a lot of setbacks and had to dig deep to get anything working. But that’s where our resilience kicked in. We took a short break, grabbed some pizza, laughed at ourselves a little, brainstormed how we can execute our idea and pushed through. By the night hours, we finally were able to get the algorithm working and between small naps and the early morning hours, we had a working demo and a polished pitch.

In the end, out of almost 25 teams, we made it to the final pitch round, standing among the top teams. Although in the end we missed the podium and finished fourth with an honourable mention, we were able to reflect that it’s not about the initial idea or setbacks you are facing, but it takes extraordinary teams to execute on exceptional ideas. It’s about resilience, the ability to push through and compete. It was an incredible experience and ultimately rewarding journey that highlights how we roll up our sleeves and innovate alongside the best. Europe has top tier talent and we at redalpine are meeting these right at the beginning of their journey. 

We talk a lot about backing the next generation of European founders. This weekend, we didn’t just meet them - we were them.