I am a PhD candidate in Quantum Machine Learning at the University of Basel, working at the intersection of mathematical physics, theoretical computer science, and statistical learning theory. My research focuses on the interplay between trainability, computational complexity, and generalization: when quantum models can be trained efficiently, what makes them hard to simulate classically, and how their architectures determine statistical performance.

I approach these questions using high-dimensional probability, random matrix theory, optimization, and computational complexity theory.

I am part of the Optimization of Machine Learning Systems Group. My research is supported by the Swiss National Science Foundation and conducted in collaboration with IBM Research Zürich.

My interest in computation began with a curiosity about how software can simulate physical worlds. Outside research, I produce music across different genres and styles; some of it is available on YouTube.

We develop the Taylor surrogate by expanding quantum cost landscapes locally around Clifford circuits, building classically tractable patches and tracking the transition to super-polynomial resource scaling for the surrogate. We connect this complexity analysis to gradient scalability, alongside our results on the Linear Clifford Encoder.

We characterize double descent with correlated features and translate the theory to quantum extreme learning machines. Encoding strength, input reuploads, and measurement shots provide ways to reshape feature covariance and stabilize learning near interpolation.