About Me
I work at the crossroads of robotics, safety, and intelligent decision-making — building agents that can learn, adapt, and act reliably even when faced with adversarial inputs. From training DRL agents in simulation to testing them on real vehicles equipped with LiDAR, GNSS, and camera systems, I aim to bridge the gap between research and reality in embodied AI.
Research
The Evolution of Criticality in Deep Reinforcement Learning
Investigates how criticality — a measure of an agent's sensitivity to in critical states — evolves throughout the training of deep reinforcement learning agents.
Advancing Robustness in Deep Reinforcement Learning with an Ensemble Defense Approach
Proposes an ensemble-based defense framework that improves the adversarial robustness of DRL agents in autonomous driving scenarios.
UrbanIng-V2X: A Large-Scale Multi-Vehicle, Multi-Infrastructure Dataset Across Multiple Intersections for Cooperative Perception
A large-scale cooperative perception dataset captured at three intelligent urban intersections in Ingolstadt, Germany, enabling research into V2I and V2V communication.
Towards Robust Agents: A Survey of Adversarial Attacks and Defenses in Deep Reinforcement Learning
A comprehensive survey of adversarial attack strategies and defense mechanisms in deep reinforcement learning, with a focus on safety-critical applications.
Real-Time Evaluation of Autonomous Systems under Adversarial Attacks
Presents a real-time evaluation framework for assessing the performance of autonomous systems when subjected to adversarial perturbations.
RoAd-RL: A Unified Library and Benchmark for Robust Adversarial Reinforcement Learning
An open-source benchmarking framework providing unified abstractions for policies, attacks, defenses, and robustness metrics in adversarial RL. Evaluates DQN, PPO, and SAC under 192 attack-defense configurations across LunarLander and Highway-v0.
Talks
Adversarial Attacks in Deep Reinforcement Learning: A Call for Robust Defenses
Jan 30, 2025 · Munich, Germany
Explored how adversarial perturbations target DRL systems, revealing critical vulnerabilities that compromise performance and reliability. Covered key attack strategies, their implications for autonomous systems, and the emerging need for robust, scalable defense mechanisms.
How Robust Is Your Self-Driving AI?
A deep dive into the robustness challenges facing self-driving AI systems, examining adversarial threats and the path toward reliable autonomous driving.
Projects
Experience
AI Researcher – Team Lead
Technische Hochschule Ingolstadt
- Developed and led a full autonomous driving stack with ROS2 for real-car deployment
- Researching adversarial attacks and defenses in Deep Reinforcement Learning (DRL)
- Published at ICAART'25, ITSC'25, NeurIPS'25, IV'26, and IEEE ACCESS
AI Robotics Engineer
Franka Emika GmbH
- Designed learning engine and modular C++ test framework for Franka robotic arms
- Developed safe ML pipelines and guided safety practices across the team
Robotics Software Engineer
Quantum Systems GmbH
- Built path planning GUI with PyQt for UAV landing operations
- Developed drone task automation stack, unit tests, and CI/CD tooling
Junior Robotics Engineer
ARE23 GmbH
- ROS-based planning and perception using OpenCV and Keras
- Built AWS workflows, web deployment pipelines, and internal testing suites
Master Thesis Student
Innok Robotics GmbH
- Developed robot diagnostics tool and visualization pipeline in Python and ROS
Senior Analyst
Flex India Pvt. Ltd
- Led material planning and lean automation initiatives
- Applied analytics (Python, R, SQL, Tableau) to optimize supply chain operations
Education
Ph.D. Artificial Intelligence (Dr. rer. nat.)
Technische Hochschule Ingolstadt
M.Eng. Mechatronics & Cyber Physical Systems
Technische Hochschule Deggendorf
B.Eng. Mechanical Engineering
Anna University