Experience

ML Engineer — Reinforcement Learning

• Led the development of Predictive Safety Curricula (PSC; arXiv preprint, 2026), using a distributional safety critic to adaptively allocate reinforcement-learning training experience toward conditions with elevated predicted risk.
• Evaluated PSC across controlled simulation, production training stacks, and physical robots; reduced shank-collision incidence by 63% relative to a learning-progress curriculum across three matched ANYmal-D training seeds.
• Developed and deployed reinforcement-learning locomotion policies for industrial quadrupeds, with GPU-scale simulation, multi-seed robustness studies, and sim-to-real evaluation.
• Secured and led a EuroHPC allocation of up to 50,000 H100 GPU-hours for RL robustness and curriculum-learning experiments.

Research Assistant

• Developed inverse-reinforcement-learning methods for causally invariant reward functions from heterogeneous demonstrations.
• Developed generalized sliced-Wasserstein distribution-matching objectives for imitation learning under dynamics variation.
• Built an RL benchmark for autonomous arthroscopic navigation and manipulation in simulation.
• Supported teaching in advanced machine learning, learning theory, and game theory.

Research Assistant

• Extended sequence-to-sequence models with structured variational inference.
• Investigated hyperbolic Wasserstein autoencoders for hierarchical representations.

Electrical Engineer

• Developed energy-harvesting wireless sensor networks for monitoring electrical machines.

PhD Computer Science (ML/AI)

Doctoral thesis: “Reinforcement learning from demonstrations: methods and applications in digital twin simulations.”

• Developed methods for causally invariant reward learning and imitation learning from heterogeneous demonstrations.
• Built reinforcement learning environments and evaluation methods for arthroscopic surgical training.
• Doctoral thesis (PDF)

MSc Electrical Engineering and Information Technology

Master thesis: “Semi-Supervised Learning using Canonically Correlated Convolutional Networks.”

• Studied multiview semi-supervised learning in convolutional networks using canonical correlations from unlabeled data.
• Thesis-related paper (2016): Semi-Supervised Learning with Correlated Convolutional Networks (PDF).

BSc Electrical Engineering

Control theory, embedded systems, and electrical engineering.