27.04.2026 aktualisiert
100 % verfügbar
Computational Materials Science Engineer | Simulation, Machine Learning & Data-Driven Engineering
Renningen, Deutschland
Weltweit
Master of Science in Computational Materials ScienceÜber mich
M.Sc. Computational Materials Science graduate with Mechanical Engineering background. Skilled in Python, ML, data analysis, fatigue assessment, FEM, and engineering software tools. Experienced in PyTorch models, signal analysis, dashboards, automation, and technical data workflows.
Skills
HTMLJavaScriptAbaqusData AnalysisKünstliche Neurale NetzwerkeBash ShellC++CSSLinuxFinite-Elemente-MethodenR (Programmiersprache)Statistische HypothesentestsPythonLineare RegressionMATLAB
I specialize in Python-based data analysis, scientific computing, machine learning, and engineering-data workflows, with a background in Computational Materials Science and Mechanical Engineering. My technical strengths include Python, NumPy, pandas, Matplotlib, Plotly, PyTorch, scikit-learn, MATLAB, C++, Bash, R, HTML/CSS/JavaScript, Git, Linux, VS Code, and Jupyter Notebook.
During my master’s thesis at Robert Bosch, I developed Python/PyTorch-based machine-learning models for metal fatigue assessment. My work included structured dataset generation, model development, hyperparameter optimization, robustness testing under noise, and RMSE-based validation. I worked with engineering calculation logic based on fatigue-assessment methods and evaluated model accuracy, generalization, and interpretability.
During my internship at Purem by Eberspächer, I built engineering-data tools for practical technical workflows. I developed a web-based analytics tool for fatigue S/N database exploration with dynamic tables, filtering, visualization, and export features. Separately, I built a Python/Tkinter desktop tool for measurement-signal analysis, including CSV import, multi-channel signal inspection, segmentation, statistical feature extraction, histogram comparison, and CSV export.
I can support projects involving Python automation, CSV/data preprocessing, technical data visualization, dashboards, engineering data analysis, machine-learning prototypes, fatigue/materials data analysis, signal analysis, numerical methods, and scientific computing. My focus is on clean, reliable, and well-documented solutions that are useful for real engineering and research workflows.
Sprachen
DeutschGrundkenntnisseEnglischverhandlungssicher
Projekthistorie
Thesis: Development of a new data-driven method for metal fatigue assessment
• Developed a machine-learning-based framework for metal fatigue assessment using Kolmogorov-Arnold Neural Networks (KANs) in Python and PyTorch to learn nonlinear relationships embedded in analytical fatigue calculations based on the Forschungskuratorium Maschinenbau (FKM) Guideline.
• Constructed staged datasets of increasing complexity, including synthetic benchmarks, FKM-based notch influence modeling, hierarchical design-factor chains, and fatigue limit prediction to systematically evaluate model behavior.
• Implemented hierarchical (concatenated) KAN architectures to emulate the multi-step dependency structure of FKM fatigue assessment, enabling structured learning of intermediate fatigue factors.
• Performed hyperparameter optimization (TPE), evaluated robustness under controlled Gaussian noise, and applied pruning and symbolification (for the fatigue-limit model) to assess interpretability and generalization using RMSE-based validation.
• Systematically compared model behavior across staged problem settings to study generalization, interpretability, and suitability of structured neural architectures for data-driven engineering prediction tasks.
• Built an internal web-based analytics tool for S/ N fatigue database exploration, including dynamic table rendering, column search, and multi-select filtering to support faster engineering data lookup.
• Implemented shareable, reproducible filtering in the web tool by syncing UI filters into URL query parameters and persisting state via localStorage.
• Developed a statistics dashboard with S/ N curves and lognormal failure probability (CDF/ PDF) parts-at-risk visualization, plus exports for engineering workflows (PDF + simulation-ready text output).
• Built a Python/ Tkinter desktop tool for signal-based analysis of engineering measurement data, including CSV import, time-axis generation from sampling frequency, and interactive multi-channel signal inspection.
• Added slider-based segmentation to the Python tool and computed segment-wise statistical features including RMS, mean, variance, skewness, and kurtosis, along with stacked-histogram comparison and CSV export.