Graduation Awards awarded during Opening Academic Year

As every year, UT awarded graduation awards to one master's student from each faculty. The graduation award consists of a cash prize of €1,000 for the best master's theses of the past year.

Faculty of BMS – Marco Luis Ochoa Barnuevo

Marco Ochoa Barnuevo did his master's in Industrial Engineering and Management at the BMS faculty, in the Production and Logistics Management track. His research addresses a question at the heart of ASML's supply chain: what do you produce, and when? It is a complex problem, because an ASML machine consists of hundreds of thousands of components that compete for the same production capacity. Produce too much of one part and inventory costs pile up. Produce too little and demand cannot be met.

Ochoa Barnuevo designed a decision framework that shows planners what is in stock at any given moment and what to do the moment a shortage hits. His method improves planning and reduces inventories by 30%. On top of that he built in a reinforcement learning layer, a form of AI that learns from experience. That layer picked up patterns the framework itself did not see and improved the decisions further.

Faculty of EEMCS – Jelle Idzenga

Jelle Idzenga studied Robotics and worked on the camera pill: a swallowable capsule with a camera, used in colorectal cancer screening. For the patient it is more comfortable than a conventional colonoscopy. The drawback is that the pill drifts passively with the movements of the bowel. The clinician cannot steer it, so abnormalities are missed more often and the capsule occasionally gets stuck.

Idzenga built a system that tracks and steers the pill at the same time. A board of magnetic sensors under the patient works out where the capsule is. Above the patient, a permanent magnet on a robot arm moves it. The core of the problem is the interference that magnet creates in the sensors. Idzenga found a way to cancel it out, and only then can you measure and steer at once. In tests on artificial and real bowel tissue the localisation worked, and he was able to turn the capsule on command. This is a first design, not a device that will appear in hospitals tomorrow.

Faculty of ET – Alexey Chechin

During his Master’s degree in Construction Management & Engineering, Alexey Chechin studied the urban heat island effect. Concrete and asphalt trap heat, which can leave a city centre up to seven degrees warmer than the countryside around it. In the United States, extreme heat claims more lives than floods, hurricanes and tornadoes combined. Urban planners use machine learning to predict where a city overheats, but those models treat every street as an isolated island, while in reality heat flows from one street to the next.

Chechin taught the models to read the city as a single connected network, in which the temperature of every street is shaped by its neighbours. He tested the approach on Apeldoorn, Rotterdam and Montreal. That change alone made the predictions 150% more accurate. Sharper maps show planners where a row of trees or a pond will genuinely cool a neighbourhood. That matters most in the fast-growing cities of the Global South, where climate change will hit hardest.

Faculty of ITC – Houri Tayarzadeh

Houri Tayarzadeh studied Geo-Information Science and Earth Observation and asked a question map-makers rarely ask: what if you could hear a city? Cities are almost always analysed by eye, through maps, diagrams and 3D models. Tayarzadeh looked at sonification as a way of adding to that. Sonification is the translation of data into sound.

Using open data for Amsterdam, she translated four characteristics of the city into both image and sound: building age, building height, street pattern and land use. Colour, shape and size on one side, tempo, register, timbre and harmony on the other. She built a web application around it in which the same data can be seen, heard, or both at once. In a test with 25 participants, visuals alone were fastest. But the combination of image and sound produced the highest accuracy, and cost participants the least mental effort. Those with a stronger musical ear read the sonified data noticeably better. Sound does not replace the map, then, but it takes weight off the eye, and it could open the data up to people who are blind or partially sighted.

Faculty of S&T – Tijmen Sijtsma

Tijmen Sijtsma studied Health Sciences. The population is ageing and demand for care is growing, while the healthcare workforce is not expected to grow with it. Intensive care units feel that first. Rosters are built on the expected number of patients, but new admissions arrive irregularly and are hard to predict.

Sijtsma saw that one part of that inflow is predictable: planned surgery is in the diary weeks ahead. He developed a machine learning model that uses data already recorded before an operation to estimate which patients will need an intensive care bed afterwards. That is harder than it sounds, because only a small minority of operations end that way and the available data is not always complete. Sijtsma adapted the model to cope with both, and achieved results that stand up against comparable models in the literature. Hospitals could use it to see their staffing needs coming sooner.