CV

Contact Information

Name Alexandru Crăciun
Professional Title Doctoral Student
Email a.craciun@tum.de
Location Boltzmannstr. 3, Garching bei München, Bavaria 85748

Professional Summary

Doctoral student and researcher at the intersection of geometry, topology, and optimization. Investigates how results in pure mathematics can be used to understand neural network training dynamics and fundamental limits.

Experience

  • 2024 - present

    Munich, Germany

    Research Assistant
    Technical University of Munich
    Conducting research on the theoretical grounding of neural network optimization.
    • Supervision of students (Bachelor & Master, Theses & Projects)
    • Teaching Assistant for ‘Discrete Probability Theory’ and ‘Statistical Foundations of Learning’
  • 2026 - 2026

    Chennai, India

    PhD Research Stay (Upcoming)
    Indian Institute of Technology Madras
    Investigating the interplay between feature learning and training dynamics within unsupervised learning regimes.
  • 2021 - 2021

    Romania

    A.I. Research Student
    Bosch Romania
    Deployed code to optimize object recognition for self-driving capability.
  • 2020 - 2020

    Magurele, Romania

    Undergraduate Summer Internship
    Horia Hulubei National Institute of Physics and Nuclear Engineering
    Studied Quantum Information Theory (Nielsen & Chuang). Participated in a Quantum Machine Learning hackathon.
  • 2019 - 2019

    Dubna, Russia

    Undergraduate Summer Internship
    Joint Institute for Nuclear Research
    Analyzed cosmological datasets to estimate the Hubble constant. Developed numerical simulations for universe expansion trajectories.

Education

  • 2024 - present

    Munich, Germany

    Doctoral Student
    Technical University of Munich
    Theoretical Foundations of Artificial Intelligence
    • Advisor: Prof. Debarghya Ghoshdastidar
    • Mentor: Prof. Christian Kuehn
  • 2021 - 2024

    Munich, Germany

    Master's Student
    Ludwig Maximilian University
    Theoretical and Mathematical Physics
    • Thesis: On the Stability of Gradient Descent for Large Learning Rate
  • 2018 - 2021

    Bucharest, Romania

    Bachelor's Student
    University of Bucharest
    Physics
    • Valedictorian (Rank 1)

Awards

  • 2021
    DAAD Study Scholarship for Graduates of All Disciplines
    German Academic Exchange Service (DAAD)

    Full scholarship awarded for MSc studies.

Skills

Programming (Advanced): Python (PyTorch), Julia (Flux), Lean, C
Tools (Advanced): LaTeX, Git, Mathematica

Interests

Research: Optimization Dynamics, Topological Data Analysis, Theoretical Deep Learning, Geometry, Topology, Algebra

References

  • Prof. Debarghya Ghoshdastidar

    PhD Advisor, TUM School of Computation, Information, and Technology. ghoshdas@cit.tum.de

  • Prof. Christian Kuehn

    PhD Mentor, TUM Faculty of Mathematics. ckuehn@ma.tum.de