About me

About Me

I’m a Machine Learning graduate student at the University of Tübingen, currently conducting research at the intersection of reinforcement learning and Bayesian inference. My work focuses on developing principled methods that bridge probabilistic reasoning with sequential decision-making.

Currently pursuing my Master’s in Machine Learning with an Erasmus exchange at Politecnico di Milano, I explore how probabilistic frameworks can enhance reinforcement learning algorithms and how RL methods can improve Bayesian inference procedures. My research is grounded in rigorous theoretical foundations while maintaining a strong emphasis on practical implementations in real-world systems.

Current Research Projects

• SPARK: Augmenting stochastic policies with non-diagonal Gaussian distributions for more expressive action sampling in continuous control
• Generalized Simulation Based Inference: Extending GBI-ACE with guided diffusion models for improved posterior approximation
• Humanoid VLA Models: Applying Vision-Language-Action models to humanoid robotics with probabilistic policy frameworks
• Distributed Multi-Agent RL: Developing asynchronous learning methods for optimal transport problems in train scheduling

Technical Background

My research leverages a comprehensive technical foundation:

• Probabilistic Methods: Bayesian inference, variational inference, generative models (VAEs, diffusion models), Gaussian processes
• Reinforcement Learning: Policy gradient methods (SAC, PPO), model-based RL, multi-agent systems, exploration strategies
• Robotics: Vision-guided control, imitation learning, manipulation, quadrupedal and humanoid locomotion
• Tools & Frameworks: PyTorch/Lightning, Gymnasium, Stable-Baselines3, ROS2, probabilistic programming

Selected Experience

Research Assistant - Autonomous Learning Group, University of Tübingen
Working on Vision-Language-Action models for humanoid robotics under Prof. Georg Martius

Research Intern - Fraunhofer Italia
Developed vision-guided imitation learning pipeline combining Task-Parameterized Gaussian Mixture Models with trajectory optimization

Bachelor Thesis - IMBIT Lab, University of Freiburg
Expanding Action Space in Reinforcement Learning through Latent Models: Integrated Soft Actor-Critic with variational autoencoders for enhanced policy learning

ML Lab Project - University of Tübingen
Extended simulation-based inference methods with amortized cost estimation using guided diffusion models under Prof. Jakob Macke

Academic Service

• Organizer: Machine Learning Reading Club, Politecnico di Milano (Polimi Data Scientists)
• Organizational Staff: European Workshop on Reinforcement Learning (EWRL 18), Tübingen

Beyond Research

When not developing probabilistic learning algorithms, I pursue competitive archery (bronze medal at European Championships 2018), long-distance running, and photography. These activities remind me of the importance of patience, precision, and perspective—qualities that translate well into research.

I’m always interested in discussing research collaborations, particularly projects that combine probabilistic reasoning with decision-making under uncertainty. Feel free to reach out via email or connect on GitHub.