Complex work,
made usable.

DroneAtlas brings RGB, thermal, multispectral, LiDAR, and elevation layers into one browser map. A team from the Netherlands eScience Center and the University of Amsterdam's 4D Research Lab developed it for archaeological research, with a workflow that can be reused for other drone datasets.

Cloud Optimized GeoTIFFs load in the browser for layered map inspection. Machine-learning-supported anomaly exploration can draw attention to unusual areas, while expert labels remain interpretation rather than automated truth. A dedicated 2D/3D presentation route and static deployment keep the maps easy to share.

DSA Case Law Tracker is a public, open-access prototype for browsing private enforcement cases under the EU Digital Services Act. It is being prepared for the Institute for Information Law at the University of Amsterdam and the DSA Observatory as part of their 2026 private enforcement research work.

The public interface turns structured case records into searchable, filterable cards and detail pages with jurisdictions, parties, legal classifications, timelines, and source references. A PocketBase editorial workflow separates review from publication, so researchers can prepare and check records before they appear in the public tracker.

Modern football can track each player as a moving 3D skeleton rather than a single dot. FIFA calls the technology skeletal tracking. It supports tactical and performance analysis as well as officiating; at the 2026 World Cup, player tracking was also combined with AI-enabled 3D player avatars for semi-automated offside replays.

At the 14-hour Ajax Hackathon, our multidisciplinary team used TF15 data from Ajax vs Fortuna Sittard to explore another use. We reconstructed 22 players from 21 body points per player at 25 fps, then built 3D playback, player POV, head-to-ball angle, scan rate, and fixation analysis. Our team won an award.

Data Storytelling reads project metadata from the Research Software Directory and presents it in four audience modes: communications, academic, leadership, and quick overview. Projects can also be browsed as cards in a Threlte 3D field.

After an audience mode is selected, Google Gemini streams a narrative from the project record. SvelteKit and Threlte, Three.js for Svelte, keep the source metadata, 3D browsing, and generated text in one browser experience.

NLeSC Portfolio is a public catalogue of Netherlands eScience Center projects, designed for editorial browsing. Project cards and filters make it easy to move through the work and open fuller project stories.

Each story links back to its people, software, and research. Lightweight authoring lets teams add material without taking on a heavy content management process.

PlanEO is a static browser map for Earth-observation data stored as Cloud Optimized GeoTIFFs. MapLibre and Svelte render the files without a custom geospatial backend.

A natural-language agent works beside the map controls, giving researchers another way to query the loaded data. The deployment remains static while COGs provide the map layers directly.

COLLaiTE is the research project behind Collens, a web tool for scholars comparing annotated textual variants. It supports TEI/XML texts, side-by-side comparison, annotation-aware alignment, and offline research sessions.

Machine-learning assistance stays inside the reading flow. Scholars drag documents in, compare variants, inspect their alignment, and keep annotations visible while they read.

SmartPlay is a Wear OS research app for studying children's play. It records real-time GPS, motion, heart-rate, and environmental signals alongside input entered by the children themselves.

The watch keeps participant input brief and glanceable. Data stays on the device and can be extracted offline for privacy-conscious fieldwork.

FAIVOR is a FAIR validation platform for medical machine-learning systems. It supports documented validation workflows with provenance, protocols, model outputs, and explainability material for reviewers.

This browser viewer, created with the Ruisdael Observatory at TU Delft, renders measured cloud points and DALES simulation output as 3D volumes.

Compressed Zarr data feeds timeline playback, a map, and cloud-point inspection. Researchers can move through changes in the boundary layer without leaving the browser.

Carbon Budget Explorer helps policymakers compare climate pathways, remaining carbon budgets, scenario assumptions, and uncertainty. The redesign keeps those differences visible while making navigation and comparisons easier to follow.

Work covered a new logotype, clearer navigation, side-by-side comparison views, and a tighter information hierarchy.

Notidian is a local-first workspace in development for notes, documents, drawings, and media. It is built around user-owned files, fast local interaction, private sync, and a canvas/editor intended to extend into diagrams, presentations, and multimedia work.

IdeasDiamond organizes ideas as visual cards inside named spaces and categories. People can keep an organization private, invite feedback, or publish selected collections on public pages.

Each space can be customized with its own name, imagery, and structure. The same card format works for personal collections and organization pages.

Via Appia is a 3D point-cloud view of the Roman Appian Way for desktop, mobile, and VR. Presented at Het Valkhof Museum in Nijmegen, it connects viewpoints and story pages to years of academic and artistic research.

NuxtJS, ThreeJS, and PotreeJS render the scans. A Git-based content management system supplies the point views and story content.

Eshmun is a standalone medical image analysis and annotation viewer built for the FEDMix research project: Fusible Evolutionary Deep Neural Network Mixture Learning. The application gives researchers a cross-platform desktop environment for loading clinical scans, inspecting model output, and drawing or reviewing anatomical regions directly on top of medical image slices.

I worked on the research-software product layer, building the clinical imaging flow as a C++/Qt/VTK application for macOS, Windows, and Linux. Researchers can inspect grayscale scans, review color overlays, edit contours, and compare panels before deciding what to make of an experimental model output.

Redesigned with Lieke de Boer and the NL-RSE community, this website brings events, resources, community updates, and the RSE feed into a clear editorial structure.

NuxtJS and Nuxt Content generate the static site. Reusable content patterns keep volunteer updates manageable, while custom animations give each section its own pace.

The EU Horizon 2020 RECEIPT project uses climate storylines to explain risks, impacts, and adaptation challenges to stakeholders. Connected scenes combine maps, 3D visuals, explanatory panels, and supporting resources.

I developed the UI from scratch. Scene continuity preserves place and story position as people move between those materials, so a discussion can follow the storyline without losing its thread.

As Lead Software Engineer and Designer, I created MisterGreen Electric's logo, palette, typography, and design system, then built those decisions into mistergreen.nl.

FocusDiamond pairs 55-minute work blocks with short breaks, a distraction capture list, and a fixed daily structure.

During a session, a capture list holds distractions for later. Named block patterns show focused time and breaks before work starts.

This 2017 Progressive Web App prototype uses service workers, mobile installability, responsive layouts, and hardware access. HTML, React, CSS, and AR.js keep desktop, tablet, and mobile on one codebase.

The browser app works offline, installs on a device, and opens augmented-reality scenes without a native app.

Illustrated Portfolio collects early work through 2014: web design, native app interfaces, UX concepts, visual systems, and freelance and client experiments.

The original PDF preserves the raw illustrated archive as it was assembled at the time.

LeapLearn uses Leap Motion to let people train three-dimensional hand gestures, connect them to system actions, and reuse them without writing low-level recognition code.

The design applies End User Development and Programming by Example. A modified $1 recognizer interprets each gesture while the screens show the recorded 3D motion for review.