Graduation Conference | Tuesday, July 8th

Software AI diagnostic tool

Sebastian S.B.A. Bokkestijn Grandez Title: AI diagnostic tool Description:

The goal of this project is to look into using deep learning for spotting anomalies to improve diagnostic accuracy and speed up resolutions. This project ties into DAF's larger goals of using AI to upgrade after-sales services and improve fleet reliability.

Location:	Eindhoven
Date:	Tuesday 8 July 2025
Time:	09:00
Room:	R10 2.18

Company:	DAF
Main IT area:	Software
IT area:	AI Engineering
Language:	English

Software Cloud monitoring

Cas C. Esselink Title: Cloud monitoring Description:

Gedurende mijn stage bij PinkRoccade Local Government heb ik gewerkt aan het ontwikkelen van een monitoringsoplossing voor IZaaksuite. IZaaksuite is een van meerdere producten van PinkRoccade. Het is een zakensysteem voor gemeenten. Dit product bestaat uit verschillende onderdelen verspreid over vijf in de cloud gehoste servers. De service provider biedt onvoldoende monitoring aan waardoor er problemen zijn op het gebied van betrouwbaarheid en beschikbaarheid. Ik heb onderzocht welke onderdelen van IZaaksuite een rol spelen in dit probleem. Per onderdeel heb ik onderzocht, welke informatie kan worden gemonitord om in de toekomst problemen te kunnen voorkomen. Deze verschillende informatie bronnen heb ik geïntegreerd in een centrale monitoringsoplossing. Door middel van dashboards wordt de huidige status van IZaaksuite inzichtelijk gemaakt, om problemen in de toekomst sneller op te kunnen lossen of voorkomen.

Location:	Tilburg
Date:	Tuesday 8 July 2025
Time:	09:00
Room:	P8 1.44

Company:	PinkRoccade
Main IT area:	Software
IT area:	N/A
Language:	Dutch

Software Agri Spectacles: Real-Time Cucumber Weight Estimation with Computer Vision for Augmented Reality Applications

Jelle J.C.C. Maas Title: Agri Spectacles: Real-Time Cucumber Weight Estimation with Computer Vision for Augmented Reality Applications Description:

Harvesting accounts for a major share of labour costs in cucumber production, yet current weight estimation relies on visual assessment and intuition, making it prone to error and inefficiency. This is especially challenging as the task is typically performed by seasonal workers, who require time to develop the necessary intuition and muscle memory for this task. This study presents a real-time computer vision algorithm for estimating cucumber weight, combining stereo depth estimation, semantic segmentation, and geometric feature extraction. The algorithm uses PIDNet-S for segmentation and LightStereo-L for stereo depth estimation, achieving a Mean Absolute Error (MAE) of 13.5 g (3.8%) at 17.1 FPS, while LightStereo-S reached 27.5 FPS with a slightly higher MAE of 19.9 g (5.5%). Weight is estimated by first calculating fruit volume and then applying a known density. These results demonstrate that accurate, real-time cucumber weight estimation on resource-constrained AR devices is feasible, supporting the deployment of smart glasses to improve accuracy during harvesting, reduce profit losses, and increase worker productivity.

Location:	Eindhoven
Date:	Tuesday 8 July 2025
Time:	10:00
Room:	R10 4.12

Company:	VBTI
Main IT area:	Software
IT area:	N/A
Language:	English

Organisational Processes BioCleanse: A Preprocessing Pipeline for Qualitative Biodiversity Images

Nikita N. Gavrilov Title: BioCleanse: A Preprocessing Pipeline for Qualitative Biodiversity Images Description:

During my graduation internship at Fontys University, I developed BIOCLEANSE, a data preprocessing pipeline that enhances the quality of biodiversity image datasets for AI model training. Focusing on iNaturalist's Research Grade data, which is widely used in biodiversity ML research but lacks standardized image quality control, BIOCLEANSE automatically filters images based on three key issues: composition (blur, exposure), human presence (body parts), and other species presence. By integrating both traditional and AI-based image quality assessment techniques, the pipeline significantly reduces noise and improves species identification accuracy. I evaluated its impact using new benchmarks with visually similar plant species, showing improved zero-shot learning performance. Released as open-source software, BIOCLEANSE aims to support the development of more robust, real-world-ready biodiversity AI models and invites further application across diverse taxa.

Location: Eindhoven
Date: 8 July 2025
Company: Fontys University of Applied Sciences
HBO-Layer: Organization Processes
IT-area: AI

Location:	Eindhoven
Date:	Tuesday 8 July 2025
Time:	11:00
Room:	R10 2.42

Company:	Fontys University of Applied Sciences, Naturalis Biodiversity
Main IT area:	Organisational Processes
IT area:	N/A
Language:	English

Software Softmedia Knowledge Management System

Stefan S. de Boer Title: Softmedia Knowledge Management System Description:

During my graduation at Softmedia, i designed and developed a Knowledge Management System that helps Softmedia solve ongoing challenges regarding scattered project and client information. The goal was to replace the outdated and rarely used intranet with a practical, user-friendly platform tailored to Softmedia's workflow. The platform centralises important project and client information that is scattered across multiple third party tools. This helps developers and management find the information they need faster, and improves collaboration.

Location:	Eindhoven
Date:	Tuesday 8 July 2025
Time:	11:00
Room:	R10 2.43

Company:	Softmedia B.V
Main IT area:	Software
IT area:	Fullstack Development
Language:	Dutch

Software Next gen ticketing template application

Pepijn P.J.H. van Loenhout Title: Next gen ticketing template application Description:

Gedurende mijn afstudeerstage bij het CM.com ticketing team ben ik bezig geweest met het migreren van een feature in een kaartverkoop SaaS oplossing. Momenteel is er een migratie gaande waarbij er twee kaartverkoop platformen tegelijk online staan, waarvan er één zich in een beta fase verkeert. Dit platform mist dan ook nog bepaalde features, waaronder (voorheen) het kunnen ontwerpen van de kaartjes op A4 formaat. Deze feature zit nu in de applicatie verwerkt door middel van een full-stack oplossing. Hiervoor is een ondersteunend CMS gebouwd waarbij de kaartontwerpen allemaal beheerd kunnen worden en is er functionaliteit gebouwd om de tickets in HTML en PDF formaat op te leveren aan klanten of in te kunnen zien tijdens het beheren van de ticket designs. Bij deze implementatie is rekening gehouden met het onderhoudbaar herbouwen van de feature, het reduceren van data complexiteit en de compatibiliteit met andere systemen in het ticketing eco systeem. Daarnaast was er ook een mogelijkheid nodig om ticket ontwerpen van het oudere platform te migreren naar het nieuwe platform, met een andere data structuur. Verder is er ook nog onderzoek gedaan naar mogelijke implementaties van roterende barcodes om kaarten beter te beveiligen tegen vervalsing.

Location:	Tilburg
Date:	Tuesday 8 July 2025
Time:	11:00
Room:	P8 1.44

Company:	CM.com
Main IT area:	Software
IT area:	Fullstack Development
Language:	Dutch

Software Migration of .NET Framework applications to a container-based infrastructure.

Teun T.A. Mos Title: Migration of .NET Framework applications to a container-based infrastructure. Description:

Zenya develops software for quality and risk management in healthcare (things like document management and audits). Part of their applications are built with the .NET Framework, which makes it challenging to host them in a scalable and modern way in the cloud.

This project focuses on researching and improving the containerization, scalability, and deployment of these legacy applications. Windows Containers have been chosen as the solution, and a proof-of-concept has been set up where the existing app services have been converted to containers.

Additionally, I have also investigated how the more modern Vue projects can be decoupled from the .NET Framework app service. The application of Infrastructure as Code is also being considered.

The goal of the project is to find a modern, future-proof way to manage both the old and new applications.

Location:	Eindhoven
Date:	Tuesday 8 July 2025
Time:	12:00
Room:	R10 4.12

Company:	Zenya
Main IT area:	Software
IT area:	Test automation CI/CD
Language:	Dutch

Software Cloud Cost Optimization for the Glass Platform

Nur N. Nechushtan Stamper Title: Cloud Cost Optimization for the Glass Platform Description:

Financial technology leader Ortec Finance faces significant cloud operational costs for its advanced Azure-based systems. My graduation project tackled this challenge through two approaches to optimize their Azure expenditure.
First, I investigated substantial cost-saving opportunities for the Sailfish High-Performance Computing framework. My analysis and prototyping demonstrated that migrating compute-intensive worker components to Azure Kubernetes Service (AKS) leveraging Spot VMs could reduce computational costs by up to 95% compared to the standard Azure Red Hat OpenShift (ARO) setup.
Recognizing the inherent eviction risk of Spot VMs in production, the second phase focused on building resilience. I designed and prototyped an intelligent Kubernetes Operator for ARO. This controller dynamically manages the balance between Spot and On-Demand worker nodes by reacting to real-time Azure eviction signals (via a custom agent and Prometheus) and Spot Placement Scores. It then automatically adjusts OpenShift MachineAutoscalers, ensuring optimal cost savings while maintaining cluster stability and operational trust.

Location:	Eindhoven
Date:	Tuesday 8 July 2025
Time:	12:00
Room:	R10 2.65

Company:	Ortec Finance
Main IT area:	Software
IT area:	Cloud Computing
Language:	English

Software Validatie van de agnostische werking van Evolve’s GraphQL-Federation met Medusa

Arnold A.C. Francisca Title: Validatie van de agnostische werking van Evolve’s GraphQL-Federation met Medusa Description:

Tijdens mijn afstudeerstage bij Lab Digital (Evolve) onderzocht ik of de GraphQL-Federation-laag écht platform-agnostisch is door de open-source commerce-oplossing MedusaJS te integreren. Ik bouwde een proof-of-concept dat:
• productdata én acties zoals “add to cart” via de bestaande, generieke schema’s beschikbaar stelt;
• mappers, resolvers en types zo ontwerpt dat alle Medusa-specifieke details verborgen blijven;
• succesvol een externe storefront vult, waardoor achterliggende systemen uitwisselbaar zijn.

Het resultaat is een herbruikbare integratie-aanpak met gedocumenteerde patronen, technische keuzes en valkuilen voor toekomstige koppelingen, waarmee Evolve zijn e-commerce-architectuur flexibel kan uitbreiden zonder vendor lock-in.

Location:	Eindhoven
Date:	Tuesday 8 July 2025
Time:	12:00
Room:	R10 4.74

Company:	Lab Digital
Main IT area:	Software
IT area:	Front-end Development
Language:	Dutch

Software Visual Exploration of Large Software Structures

Kiril K.I. Katsarski Title: Visual Exploration of Large Software Structures Description:

As software systems increase in size and complexity, maintaining their interpretability presents a significant challenge. Many have sought effective solutions through the use of diagrams and visualization techniques aimed at presenting software architectures in more intuitive and accessible ways. This project introduces VELSS, an application designed to dynamically analyze and visualize the hierarchy and file-to-file dependencies within large C/C++ codebases. The assignment was focused on identifying stakeholder requirements and selecting appropriate technologies to develop a functional prototype from the ground up. It involved examining existing tools and their limitations, assessing visualization methods suitable for hierarchical structures, and evaluating tools for both interactive visualizations and C/C++ dependency analysis. Based on the findings, a system was proposed and implemented, leveraging Clang tooling for dependency extraction and modern WebGL-based technologies for dynamic, interactive visualizations, combined in a developer-oriented desktop application. Iterative prototyping, guided by continuous validation with stakeholders, shaped the refinement of the product. The final implementation successfully delivers a practical, interactive solution for exploring complex software structures.

Location:	Eindhoven
Date:	Tuesday 8 July 2025
Time:	14:00
Room:	R10 2.31

Company:	Sioux Technologies B.V.
Main IT area:	Software
IT area:	Fullstack Development
Language:	English

Software Algorithmic risk assessment from passive bed sensors for pressure injury prevention in elderly care

Youssef Y. El Jaddaoui Title: Algorithmic risk assessment from passive bed sensors for pressure injury prevention in elderly care Description:

During my graduation internship at Fontys, I conducted a study that explores whether sensor data from beds in elderly care can be turned into clear and useful information to support early action against pressure injuries, also known as bedsores or decubitus. These injuries often develop when a person stays in the same position for too long or does not move enough to relieve pressure on the body. In elderly care settings, it can be difficult for staff to notice these risks in time, especially since residents are often more fragile and can develop complications more quickly. The goal of the study is to test whether patterns in daily behavior, such as time spent in bed, limited posture changes, and time with the bed in a raised position, can be used to create clear insights and risk scores that give caregivers useful information to act on and help prevent complications.
The prototype developed during the internship was built using real data from elderly nursing homes and follows clear rules based on input from care professionals. It focuses on providing daily and periodic overviews that are easy to follow and understand. These overviews show when a resident might need more attention. The tool is designed to support daily routines and does not replace the knowledge or experience of care staff. The study shows that sensor data already being collected can be used in a smarter way to improve awareness, guide decisions, and help prevent problems before they grow.

Location:	Eindhoven
Date:	Tuesday 8 July 2025
Time:	14:00
Room:	R10 2.43

Company:	Fontys Hogescholen Eindhoven Lectoraat
Main IT area:	Software
IT area:	N/A
Language:	English

Tuesday, July 8th

Eindhoven - R10

Tilburg - P8