Hey! 👋🏼

Daniel (my colleague) and I attended ScaDS.AI's summer school on Big Data and AI. Over the course of several seminars, we learned a lot from new and exciting research fields, and met lots of cool people. 🔗 Read more

In 2021 I first joined the OUTPUT team as a student and rebuilt the OUTPUT app from the ground up for iOS. After I finished my studies, I am now responsible for the app's development. For OUTPUT 2023, we made a radical choice and decided to rebuild the event‘s app in Flutter to cut our future development efforts drastically. This was only possible thanks to the talented team of four students that I had the chance to guide. Over the semester, we planned and realized the rebuild and successfully deployed our app at OUTPUT 2023. Looking forward to OUTPUT 2024! 🔗 Read more

The PrioBike open beta has just launched! 🚀 What is the PrioBike app? It's a smart navigation app for cyclists in Hamburg that provides speed recommendations to catch as many green lights as possible. Huge thanks to my dev team Charles Krüger, Daniel Jeschor, Markus Wieland, and Paul Pickhardt. 🙌 An additional thank you goes to our project coordinator, Thomas Springer, and all students who have contributed their time and effort to make our ideas come to life. 📚

PrioBike was nominated for the German Cycling Award 2023, and got the third place! 🔗 Read more

Selecting Resource-Efficient ML Models for Transport Mode Detection on Mobile Devices 🔗 Read more

I tried out stable diffusion a few days ago. Here are some AI-generated cyberpunk-themed artworks. Stable diffusion is a state-of-the-art image generation technique that utilizes an AI model. This model has been trained to learn an internal representation of human imagery. In this way, it also learned what cyberpunk looks like. Being given random noise as an input and a guidance prompt, this AI is able to create completely unseen artworks with some kind of creativity. To me, this is absolutely mind-blowing 💥

Matching Traffic Lights to Routes for Real-World Deployments of Mobile GLOSA Apps 🔗 Read more

The 3Dify app generates disparity maps for portrait images. On devices without that capability, it utilizes a machine-learning model for monocular depth interpretation. Based on the disparity maps, a technique called parallax occlusion mapping is applied to generate a 3D-like effect. Parallax occlusion mapping is utilized in video games to simulate 3D surfaces without having to generate any additional vertex geometries. The 3D process is calculated entirely in the GPU's fragment shader. 🔗 Read more

In many use cases of smartphones, such as activity recognition, we need machine learning models to solve the underlying problem. However, machine learning models can grow vastly under certain circumstances and consume many computational resources. Thus, it is crucial to consider the resource factor in the model development. In my Diploma thesis, I found for human activity recognition models that many research papers primarily focus on model accuracy and lack deeper considerations of resource consumption. I developed a 3-step engineering process for machine learning models that uses resource-enhanced grid search to identify a model configuration with the best resource-accuracy-tradeoff. In my talk at OUTPUT.DD 2022, I will outline my process and learning outcomes, and discuss why porting and engineering machine learning models for smartphones is still an ongoing problem. 🔗 Read more

I am excited to announce that I joined TU Dresden as a research associate and will work on the PrioBike project, which implements a smart velomobility app for cyclists in Hamburg. The planned time horizon is to finish my Ph.D. thesis together with the project by the end of 2024.

Together with my colleague Felix Kästner we developed a proof-of-stake blockchain in Go. The goal of this project was to learn more about blockchain - and what would be a better way than implementing one yourself? We deployed the blockchain in a small scale via Heroku, and were able to synchronize transactions to the blockchain from our mobile apps. The hardest part was getting the P2P networking right. 🔗 Read more

In this project, I wanted to challenge myself a bit more and develop a low-level camera interface that utilizes a machine-learning model for image segmentation to replace the background. The plugin can be plugged into any application since it acts as a virtual camera wrapper around the system's camera. The project is quite prototypical but was worth the laughs 😃

To aid learning for exams, I wrote my own questionnaire platform: fragenautom.at - It is available for free and provides unofficial questionnaires for the TU Dresden courses Distributed Systems, Computer Vision, Machine Learning, User Interface Engineering and more.

How loud is it right now? This app tells you. My new app offers the following features:
- Easy to use: Simply start the measurement and directly read the measured volume.
- Precisely calibrated dB (decibels) measurements, almost like a professional measuring device.
- Display of the current volume using a dynamic volume bar and a text display in dB (decibels).
- Dynamic images and live descriptions for the measurement, through which you can learn more about the different sound intensities and their potential health risks.
- Real-time audio frequency spectrum, with which you can directly see, whether the volume you measured is from high or low tones.
- Simply record the volume over any period of time and save it as a summary in the form of a diagram.
- Automatic calculation and display of the maximum, minimum and average volume in the diagram.
"Volume" shows you which environment the measured volume corresponds to and gives you additional, professionally researched information about the health risks of excessive noise levels.
With "Volume" you can also measure and display the available frequency spectrum in real time.

I developed a VVO App based on Kiliankoe's vvo library to locate tram and bus in Dresden. In the iOS app you have the following features:
- Intelligent stop search
- Real-time departure monitor
- Route search from stop to stop
In the watchOS app, you have the following features:
- Location-based stop search
- Real-time departure monitor

I supported the young author Victoria Scheer with the creation of unique covers. The fantasy-oriented books are matched by the special style we developed for this. The covers were created with 3D visualization or hand drawn and then stylized with neural networks. We used specific components of the Inception model to create a mystical mood and to already impress the reader before opening the book.

I had the opportunity to collaborate with the MDR trainees and contribute to the development of the "MDR Koalitionspoker" game. It was a fulfilling experience to work on this browser-based game, where players negotiate coalitions and advocate for their party's positions. Whether playing with real individuals or a computer opponent, the game offers an engaging political simulation. As the external programmer, I assisted in bringing the vision to life. For more information, see: https://www.mdr.de/nachrichten/sachsen/politik/landtagswahl/video-320684.html

Together with the medium-sized company "Hörner Bau und Handels-GmbH" I designed a new responsive website with a modern look. The photorealistic 3D visualization on the homepage links the various areas of expertise of the company. With the help of the search engine-optimized website, the company can now thrill with its online presence!

Together with the acoustic band Jarése, we created a soulful website to perfectly convey the music style of the band. The band members can now easily enter new concert dates on their website and even sell music through the integrated online shop.

For Optitak, I not only produced a modern and stylish website, but also designed a unique corporate design that fits seamlessly into the image of the precision measurement technology startup. In order to give the user an interactive experience, we also integrated explanations to the individual services of the company.

I had the opportunity to develop a social network for a client who wanted to stay in touch with other alumni, and had the idea to establish a platform for staying in touch even after graduation. The app features social features similar to Jodel. I implemented the app using Java (Android) and Swift (iOS), and utilized a Google Firebase backend.

I started to work as a tutor for Software Technology (seminar and lab) at TU Dresden. This job runs in parallel with my job as a software developer at INLOOP.

The Whatchapp App is an experimental whatsapp App for the Apple Watch, using the web interface of whatsapp and web scraping.

The Salut App is an App to control a Mac remotely via encrypted data transfer over bonjour, from an iPhone or iPad. Features:
- US/DE Keyboard
- Symmetrically encrypted connection between iOS device and Mac
- Wireless control of your media, presentations
- 3 different types of keyboards: Alphanumerical, Special, Control
- No bluetooth connection needed
- Password secured keyboard access
- Convenient and colorful interface
- Connect multiple devices to one mac
- Direct and secure data transfer without servers in between
- Uses the Bonjour protocol to communicate

The Vokabular app is an app to learn new languages by multiple-choice tests. I implemented this app to try and learn a bit of Spanish.

I developed a new app: NO2. With NO2 you can track current air pollution in your region. Use the integrated Map-View and analyze air pollution data from stations around your area. Save important stations to find them quickly and receive notifications for updated measurements. With NO2, you can discover which regions are polluted more, worldwide. NO2 contains the following features:
- View stations in your area and save them
- Trace statistics over a range of 30 days in the past
- Compare measurements with the recommended daily maximum
- Receive notifications for changing measurements and track daily changes
- Use the daily widget of this App to get informed on your lockscreen
- Compare measurements of different stations in many countries
NO2 supports the measurements of nitrogen dioxide (NO2), particulate matter (PM10 and PM25), sulfur oxide (SO2), ozone (O3) and carbon monoxide monoxide (CO). Please regard that not all stations support all measurements. NO2 uses the information supported by an open-source API called OpenAQ. For further information visit the OpenAQ website at https://openaq.org/.

I started to work at TU Dresden chair of software technology as a backend and frontend developer for INLOOP, an e-learning platform.

I developed a small app for brushing your teeth. With this app, you can control yourself and improve your oral hygiene. With the interactive timer, you can see exactly how long you still have to clean to get the best results and prevent tooth decay. With the aid of the interactive dents, you can easily identify which area of ​​your teeth you should clean. This app is also suitable for children!

The Pacebox app is an app to track and correct acceleration times in a car. You can use this app simply as a speedometer, but it can do much more than that:
- Acceleration measurements: track your speed and precisely measure your acceleration from 0 to 60 mph or any range and unit you want.
- Save your measurements and show them later to your friends - with a detail view of your measurements. Check your height difference, real time and height corrected time, GPS precision and much more!
- You won't need a GPS Black Box anymore with this app.
- Windshield Mode: turn your phone into an interactive heads up display simply by laying it under your windshield.

After I finished my Abitur, I didn't want to relax all the time until my Diploma in CS started in October 2016. Therefore I decided to bring my previous skills in 3D design to the table and got the opportunity to do some interior and product visualizations for GOLIAT Agentur. My biggest project was rendering all of the metal-made lockers of the company METAN for their new product catalog.