avatar

Sándor Battaglini-Fischer

PhD Candidate in Neuromorphic Computing
Funded by the Marie Skłodowska-Curie Actions POSTDIGITAL+ Network

Academic Work

I am a PhD Candidate in Neuromorphic Photonic Computing at IFISC, CSIC-UIB, funded by the European Union through the Marie Skłodowska-Curie Actions POSTDIGITAL+ Network.

IFISC Logo

Scroll down for details on my academic projects and research experience!


🧠 My PhD Project

POSTDIGITAL+ Logo Funded by EU
Brain-Inspired Computing with Photonic Oscillators
  • PhD Candidate at IFISC (Institute for Cross-Disciplinary Physics and Complex Systems), a Research Centre of Excellence between the Spanish National Research Council (CSIC) and the University of the Balearic Islands (UIB).
  • Funded by the European Union through the Marie Skłodowska-Curie Actions POSTDIGITAL+ Network.
  • My research focuses on brain-inspired computing based on photonic oscillators, aiming to transfer recent concepts in cognitive neuroscience and machine learning into the photonic domain.
  • The project combines aspects from physics (photonics), computer science/machine learning, mathematics (dynamical systems) and cognitive neuroscience to design novel computing architectures.
  • Contributes to the POSTDIGITAL+ project, addressing the end of Moore's law and the global energy crisis of AI and digital infrastructures.

Track my current research progress here →


📚 Education

For my full CV, please see here.


📑 Other Research Experience

My research experience spans interdisciplinary topics including computational science, machine learning, statistics, dynamical systems, scientific visualization, and biophysics.

Master’s Thesis (University of Amsterdam)

In my thesis titled Enhancing Computational Capabilities of High-Dimensional Dynamical Systems: Optimization of Stuart-Landau Oscillator Networks, I investigated delay-coupled networks of Stuart–Landau oscillators as trainable recurrent architectures for sequence processing. The project was carried out at IFISC, supervised by Prof. Dr. Miguel Soriano and examined by Prof. Dr. Greg Stephens. The computational model was motivated by laser dynamics near a Hopf bifurcation and aimed at bridging physics-based oscillator models with modern machine learning methods.

Key Achievements:
  • Implemented and optimized oscillator networks in JAX with end-to-end training using complex-valued gradients and IMEX integration.
  • Conducted ablation studies (showing the critical role of coupling and delay buffers in learning capacity) as well as a scaling study.
  • Analyzed learned dynamics, observing emergent sparsity, spectral properties, and role differentiation within the network.

FAILS Framework: LLM Service Analysis

I developed FAILS - A Framework for Automated Collection and Analysis of LLM Service Incidents, published at ICPE'25.

This work addresses the emerging challenge of understanding failure patterns in Large Language Model services like ChatGPT through automated data collection, analysis, and visualization.

Paper → Code →
FAILS Framework

Other Research Projects

Brain Visualization

Brain Data Visualization

Created custom visualization tools for investigating lesion and stimulation dynamics in the human brain using calcium imaging data. Based on the 2023 IEEE SciVis Contest.

PDF → Code →
Criminal Networks

Criminal Network Analysis

Contributed to Citadel, a web-based tool to visualize and interact with criminal network graphs. Research project for the Amsterdam Police Department using React, Node.js, and Cytoscape.js.

PDF → Code →
Chaotic Dynamics

Chaotic Relationship Dynamics

Explored emergent behaviors in complex relationships through "Romantic Chaos" - modeling relationship dynamics using coupled ODEs, analyzing Lyapunov exponents and bifurcation diagrams.

PDF → Code →
Cell Culture Analysis

3D Cell Culture Analysis

Bachelor's Thesis: Applied a U-Net to analyze complex biological data from 3D cell culture experiments at TUM, supervised by Prof. Dr. Andreas Bausch and Dr. Fabian Englbrecht.

PDF →