Biosketch#
July 2024
Past career#
As a computer engineering student I was fascinated by what tiny computers like microcontrollers can achieve by direct interaction with their environment. Collaborating on building robots for an annual competition, I contributed to PCB design and programming, which further ignited my passion for embedded systems. One notable project from my student years was independently implementing a system managing access for 450 students to four washing machines in my dormitory, which was embedded on the washing machines. This system replaced the paper & pen based reservation system and reduced the waiting time by one week. Experiencing the direct impact of my work was very rewarding.
During my internship in Philips research labs I designed an FPGA-based data acquisition system architecture for positron emissions tomography (PET, a molecular imaging technology) that could gather data from 10 sensors with 1 GBit/s bandwidth each. This system laid part of the groundwork for the first industrial PET devices based on solid-state sensors and culminated in my thesis.
With my experience in FPGAs I joined the avionics systems team at German Aerospace Center to create the in-house on-board computer for future satellite missions. The FPGA allowed a scalable system that could cater for a wide range of missions with various interface requirements regarding the number and the type. Our computer was launched to space in 2018 and functioned reliably in space.
Radiation in space causes bitflips in spaceborne digital systems. Parallel to my engineering work I pursued a PhD at University of Bremen to research on an alternative fault-tolerance technique based on error detection instead of the correction-based popular approach TMR. My technique saves up to 45% of the overhead that is caused by the standard triplication-based redundancy approach which laid foundations to enable area-efficient data processing for dependable spaceborne computing.
Collaborating with or working in industry is the key to reach the masses and increase the impact. After my PhD I joined a research group in Siemens who create security solutions for embedded devices. I helped security experts to realize their solutions on hardware, e.g., for securely logging operational data in trains. In another project we designed an FPGA-based secure communication infrastructure for industrial applications which resulted in two patents.
Later on, I was invited by a startup accelerator to join an entrepreneurship program, where I worked with other co-founders on ideas like energy-efficient bioinformatics data processing or automatic knowledge management on construction sites through speech recognition. My teams could not get any funding and I experienced that not only finding a technical solution for a problem is crucial but also to convince customers through networking.
After my industrial experience I got the opportunity to have a teaching professorship in a university for applied sciences, where I could share my passion as an engineer with international students. During five years I created many modern web-based and openly available learning resources. My latest project focuses on modern usage of the hardware description language (HDL) SystemVerilog and FPGA programming in general. One distinctive feature of the course is the integration of the 4F model for effective teaching — fail, flip, fix and feed. Teaching professors typically do not have significant resources for research. I could still attract about 300 k€ for my ideas in total for energy-efficient computing using FPGAs.
Even I find sharing my passion with students very rewarding I want to spend more time in tackling technical problems in my field of expertise. My new assistant professorship that I will start in September at the DTU will hopefully cater for that.
Vision#
The information and technology sector, including data centers, contributes approximately 2% of global carbon emissions, a figure comparable to the emissions generated by the aviation industry (2017, Avgerinou et al.). Given the existential threat posed by climate change, partly caused by carbon emissions, which could be increased by AI training workloads despite the rise of renewable energy sources. My five year goal is to research reconfigurable computing architectures for AI training & inference that can use less gates and energy through approximate computing. I could recently attract funding for a PhD for energy-efficient reinforcement learning.
An important part of carrying my field forward is to create openly-accessible teaching materials for people interested in my field. Even there are many high-quality open source studying materials for high-level programming, the materials regarding reconfigurable computing, especially FPGAs, are lagging behind. In the next five years I want to create a modern course about FPGA-programming that is based on open-source tools and hardware. The course will have a special focus on topics like wearable technology, arts, healthcare and sustainability which may attract more female students into my male-dominated field.