Final-year BSc student at the Technical University of Denmark
I’m in the final year of a BSc in General Engineering (Cyber Systems) at DTU, with a focus on mathematics and computation, and I’m writing my bachelor thesis on 3D neutron computed tomography at DTU Compute. Next I want to do research in applied mathematical analysis, combining mathematical models with numerical methods and machine learning.
Before and alongside the BSc I have worked part-time as a software developer for more than four years, first at Nordic Insurance Software and now at Dalux, and I write embedded software for DTU’s Formula Student team.
2023–2027
Selected courses in mathematics and computation:
| Mathematical Modelling | 12 |
|---|---|
| Introduction to Operations Research | 12 |
| Mathematics 1a and 1b | 12 |
| Advanced Engineering Mathematics 2 | 10 |
| Introduction to Numerical Algorithms | 10 |
| Statistics | 10 |
| Introduction to Partial Differential Equations | this semester |
Grades on the Danish 7-point scale, where 12 is the highest (an ECTS A) and 10 is an ECTS B.
2020–2023
Software development, systems design, multithreaded systems, data analysis and machine learning. Final project on a microservice management system, graded 12.
2026–present
PLC programming for laboratory equipment.
2025–present
I develop the vehicle control unit on an STM32 in C: CAN message dataflow and FSMD-based control logic. I also extend the car’s CAN software for telemetry and data acquisition, and built a real-time data visualization pipeline that works both locally and online.
2025–present
Production software in a large codebase, backend and frontend, in an agile team: design, implementation, testing and code review.
2022–2025
Three years of part-time full-stack development: data-driven systems and business logic.
2026–2027
Like X-ray CT, neutron tomography reconstructs the inside of an object from projections taken at many angles, here measured with neutrons. Neutrons pass easily through many metals but are strongly attenuated by hydrogen, so they can show water or oil inside a metal part, where X-rays mostly see the metal. The thesis covers the mathematical modelling of the measurements and the reconstruction, formulated as an inverse problem.
Object
Measurements
Reconstruction
Relative error
| Material | Neutrons | X-rays |
|---|---|---|
| Aluminium | 0.10 | 0.46 |
| Steel (iron) | 1.19 | 2.93 |
| Water | ≈ 3.5 | 0.17 |
Linear attenuation coefficients in cm−1, for thermal neutrons (PSI table) and 100 keV X-rays (NIST tables times density). The object is a sum of ellipses on a 128 × 128 grid, so each measurement is an exact line integral; the reconstruction uses the Ram-Lak filter and linear interpolation. The error is ‖frec − f‖2 / ‖f‖2 against the object averaged over each pixel. Images are shown on a square-root scale so that weak attenuation stays visible.
The report and code will be posted here when the thesis is finished in January 2027.
2025
The same image-processing task at three levels of abstraction. A C program finds cells in 950 × 950 microscope images through background subtraction, thresholding and repeated erosion. A 32-bit single-cycle CPU, written in Chisel with its own ten-instruction set and a Python assembler, runs an erosion program on a 20 × 20 image. Finally, a dedicated accelerator in Chisel, controlled by a finite-state machine, performs the erosion directly. Each part has automated tests.
Test image
After one erosion
Draw on the test image to change it.
2024
A device that follows how a yeast culture grows by measuring how much light passes through it. An ESP32 running MicroPython reads a light sensor, converts each reading to optical density with OD = −log10(I/I0) and a linear calibration, and sends it over WiFi to a Flask server. A Next.js web page shows the growth curve while the experiment runs.
| Time (h) | OD600 |
|---|---|
| 0 | 0.24 |
| 6 | 0.28 |
| 12 | 0.37 |
| 18 | 0.66 |
| 24 | 1.15 |
| 30 | 1.23 |
| 36 | 1.30 |
| 42 | 1.34 |
| 48 | 1.39 |
| 54 | 1.43 |
| 60 | 1.46 |
| 66 | 1.50 |
| 71.4 | 1.53 |
Readings closest to every sixth hour. All 8,564 are in the raw data file.
2025
PID-controlled solar heat collector. An Arduino keeps a solar heat collector at a target temperature by adjusting how far its shades are open with a stepper motor; the logged temperatures are plotted in Python. C++. Code
2024
ISA simulator. A Python simulator for a small instruction set, tested with our own sorting program. Group project, Cyber Systems course. Code
2024
FSMD simulator. Simulates finite-state machines with datapath described in XML; our own test case solves the climbing-stairs problem. Group project, Cyber Systems course. Code
2022
Claim reserve prediction. Machine-learning models that predict insurance claim reserves for Nordic Insurance Software: ridge regression, decision trees, random forests, gradient boosting and neural networks. Python. Code
