Doctoral researcher
Artificial intelligence for battery manufacturing. I develop machine-learning methods that connect process data, physical models, and electrochemical performance. At VARTA and TU Braunschweig, my work spans hands-on electrode development, continual learning, and process–property modelling for changing materials and operating conditions.
Industrial PhDVARTA Microbattery GmbHTU Braunschweig
VARTA Microbattery · TU Braunschweig
Developing and qualifying coating, calendering, and semi-dry processing routes for lithium-ion button cells.
Technical University of Munich
Synthesised and screened catalyst pairs for sequential epoxidation and CO₂ fixation within one framework.
Robert Bosch GmbH
Scaled ceramic paste processing from laboratory batches toward mass-production volumes.
Max Planck Institute · BASF
Established precursor routes for NCA and NMC and investigated lower-temperature lithiation strategies.
Continual learning, geometric interference, low-rank task structure, and robust adaptation under process drift.
Dry processing, extrusion, coating, calendering, and physics-aware process–structure–property models.
Impedance spectroscopy, cycling-data interpretation, and linking process conditions to cell performance.
Models forgetting as weight-interference energy under a task's feature covariance, and derives a replay-free structural controller (IGFA) that protects earlier tasks while allowing capacity to be shared where directions do not conflict.
@misc{stoerk2026interference,
title = {Interference and Retention in Continual Learning},
author = {St{\"o}rk, Julius},
year = {2026},
eprint = {2607.09202},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
doi = {10.48550/arXiv.2607.09202},
url = {https://arxiv.org/abs/2607.09202}
}A differentiable, calendering-aware Zehner–Bauer–Schlünder closure that tests whether contact-network growth or particle reorientation dominates through-plane thermal transport.
@misc{stoerk2026calendering,
title = {Capturing the Calendering U-Shape in Lithium-Ion
Electrode Thermal Conductivity},
author = {St{\"o}rk, Julius},
year = {2026},
eprint = {2607.11521},
archivePrefix = {arXiv},
primaryClass = {cond-mat.mtrl-sci},
doi = {10.48550/arXiv.2607.11521},
url = {https://arxiv.org/abs/2607.11521}
}A NeurIPS 2025 paper trains one transformer on simulated, identifiable causal worlds; new observational data goes in as context, treatment-effect posteriors come out. What holds it up, and where it breaks.
How learning a new task damages an old one—and how interference geometry suggests a structural controller.
Research code behind the continual-learning preprint: interference-attribution analysis and Colab pipelines for continual pretraining and LoRA fine-tuning of language models, plus the working notes and negative results that shaped the paper.
Notebooks behind the calendering closure: calibration and validation against the electrode data, alongside ML, PINN, and multiphysics comparisons on the same problem.
Instrument control from my M.Sc. work on automated high-throughput electrochemistry: pump and robot control for a scanning droplet cell.
Agent experiments for ARC-AGI-3: a Python game-playing agent with per-game diagnostic scripts and analysis notebooks.
Happy to hear about collaborations, data, or awkward results that do not fit the model.
julius.stoerk@gmail.com
linkedin.com/in/juliusst
github.com/j-stoerk
orcid.org/0009-0006-3519-0950