TU DELFT DREAM TEAM · 2025–2026
Epoch VI
Built through competition, applied to the world’s most important problems.
We were Epoch VI. Sixteen students from TU Delft who spent a year competing against industry professionals in international AI competitions, building AI systems for real organisations tackling real problems, and creating the infrastructure for the next generation of Dutch AI talent to do the same.
COMPETITIONS
4 entered. 2 big wins.
2nd out of 430 on a Gates Foundation challenge against industry professionals. 1st place at the ETH Zurich Datathon — our first international competition.
PROJECTS
AI where it matters.
Simulated the Dutch government’s policy process and presented it inside the Ministry. Built poacher-detection software for South African game reserves — and flew out to deploy it.
AI CUP 2026
A national competition.
There was no national student AI competition in the Netherlands, so we created one: 500+ participants, 7+ universities, a real conservation challenge — now recurring.
Competitions
We entered international AI competitions against industry teams all year. The season read like a progression — each result built on the last.
Brain Aneurysm Localization
Medical imaging · First competition
655 of 1,154
Automated detection of brain aneurysms from multimodal medical imaging — a genuinely hard computer-vision problem with messy, high-dimensional data. We placed mid-field, but left with a much clearer picture of what our pipeline was missing.
Grass Biomass Predictions
Satellite & sensor data · Sustainability
Top 7% of 3,806 teams
Predicting pasture biomass to support sustainable livestock management, against nearly 4,000 teams worldwide. We finished 257th — the lessons from the aneurysm competition were visible in the results.
Children’s Speech Recognition
Gates Foundation · DrivenData
2nd of 430 · $15,000 prize
The most competitive thing we entered all year: teaching AI to understand children’s voices — irregular rhythms, mid-sentence accent shifts, almost no training data. We finished 2nd of 430, in a field of professional ML engineers and research teams.
ETH Zurich Datathon 2026
Energy forecasting · Databricks × Axpo
1st place · CHF 1,500
A 24-hour datathon on reducing electricity waste across European grids. Our model cut projected waste by 50% — an estimated €1.7M saved every three months. Our first competition abroad as a team. We won.
ONE SEASON OF PROGRESSION
From mid-field to the podium in four competitions.
Projects
Two problems with no benchmarks and no leaderboards — just real constraints and real stakes, for the Dutch government and a South African conservation organisation.
DUTCH MINISTRY OF THE INTERIOR · 10-WEEK PROOF OF CONCEPT
VirtueleKamer — AI for the Dutch Government
The Ministry came to us with a question: can the government be replaced by AI? Not entirely, obviously. But could an AI system simulate how policy actually moves through government?
In 10 weeks we built a Policy Pipeline Simulator: LLM agents that navigate government data sources, simulate decision-making across departments, and produce coherent policy drafts — presented at the Ministry in front of 200+ officials.
SPOTS · SOUTH AFRICA · DEPLOYED IN THE FIELD
Frontline Anti-Poaching AI
Poaching remains a serious threat to rhinos in South African reserves. Drones survey vast areas, but reviewing the footage manually doesn’t scale. We built a computer-vision system that detects human presence automatically.
The data was brutal: no labels, low resolution, tiny figures, constant camera movement. We built the labelling pipeline from scratch — then travelled to South Africa to deploy the system in the actual reserves.
Wildlife & Field Trips
Africa is home to some of the most extraordinary wildlife on Earth — and some of the most vulnerable. Deploying the anti-poaching system took us out into the reserves these animals depend on: long drives between test sites, early mornings with the conservation teams, and, up close, the animals the whole project exists to protect.
FROM DELFT TO THE FIELD
Built in a workshop. Deployed in a game reserve.
AI Cup 2026
The idea came from a simple frustration: we wanted to compete against other Dutch AI students, and there was nowhere to do it. So we built a national competition — a real dataset, proper prizes, multiple universities and industry partners.
We partnered with TNO, who provided radar track data from Dutch wind farms: classify bird species from flight patterns, so operators can reduce bird collisions without unnecessary shutdowns. Halfway through the year, AIC4NL joined with the same vision — unite Dutch AI talent into a proper national benchmark. The result was AI Cup 2026.
500+
PARTICIPANTS
7+
UNIVERSITIES
€4,000
IN PRIZES
Ed. 2
IN PLANNING
The event was a 6-week competition, consisting of three high-energy in-person events:
FEB 14 · MONDAI, DELFT
Kick-off Day
Launch event. Students formed teams, got briefed on the challenge, and started building.
MAR 28
AI Futures Day
Top teams nominated. Industry workshops with Prosus, Nebius, RWE and TNO.
APR 14 · DUTCH AI CONGRESS
The Finals
On stage at the largest AI event in the Netherlands. The winning team took home €3,000.
APRIL 14 · DUTCH AI CONGRESS
The first AI Cup winners, on the national stage.
Other Moments
Besides work, we’ve had many fun moments throughout the year.
