I hold a Master’s Degree in Computer Science from the University of St Andrews, where I focused on machine learning, multi-core programming, and data-intensive systems. My master’s thesis centered on model-based reinforcement learning, exploring sample-efficient methods for decision-making and control in dynamic environments.
During my undergraduate degree in Electrical Engineering and Communication Technologies at the Technical University of Vienna, I focused on Robotics, Machine Vision and Control. I was also part of TU Wien Racing, where I led the embedded software development for the racecar’s electronic control unit and contributed to securing a top-25 place among over 200 international teams.
My interests lie in robotics, reinforcement learning, and control, with a focus on building intelligent and impactful autonomous systems.

• experimenting with ROS2
• learning rust

• Transformer-based World Model for Cloth Manipulation - Deep RL · Control · Manipulation

  Implementation of a Transformer-based state-space model (TSSM) for robotic cloth manipulation. The agent learns latent world models from fabric manipulation datasets and uses model-predictive control for planning. Built on the Agent-Arena framework for training, evaluation, and benchmarking of world-model agents on cloth flattening and other deformable-object tasks.

• Improved Change Detection in Autonomous Systems – ROS · DINO-ViT · Perception

  Integration of the CYWS-3D model into an indoor mobile robotics research project to enhance visual change detection. Combines DINO-pretrained Vision Transformers for feature extraction, SuperGlue GNNs for correspondence matching, and UNet-SCSE modules for spatial reasoning. Uses CenterNet detection with differentiable 3D warping to bypass traditional reconstruction pipelines, improving occlusion handling and geometric registration for unsupervised home-assistance scenarios.

• Autonomous Robot Hockey System – ROS · Sensor Fusion · Navigation

  A ROS-based multi-robot system enabling autonomous hockey-style gameplay. Implements LiDAR + Kinect sensor fusion for perception and potential-field motion planning for real-time navigation and obstacle avoidance. Features a state-machine architecture supporting team coordination, dynamic target selection, puck manipulation, and goal-scoring behavior on physical robots.

• SimpleSynth — Digital Modular Synthesizer – C++ · Audio DSP · GUI

  A node-based modular synthesizer for real-time audio generation and processing. Built with ImGui/ImNodes for an interactive interface and STK for audio synthesis and effects. Supports patch creation via drag-and-drop, live parameter editing, save/load functionality, and extensibility for custom modules.

Programming Languages:

Technologies and Frameworks:

Cryptography:

Mail: florian@pfleiderer.at