Research Topic: Virtual Diagnostics & Surrogate Models for Laser-Electron Accelerator OptimizationI am currently pursuing a PhD at Helmholtz-Zentrum Dresden-Rossendorf (HZDR) in the Institute of Radiation Physics, focusing on optimizing Free-Electron Laser (FEL) radiation from electron bunches generated by laser plasma acceleration (LPA). My research involves the integration of simulation codes (Elegant, Genesis, PIConGPU) into machine learning pipelines and developing digital twins for virtual diagnostics.Key aspects include:Optimizing FEL radiation through simulation-based analysis of electron bunches.Developing surrogate models and virtual diagnostics for experimental LPA-FEL setups.Collaborating closely with experimental teams to identify promising parameters for exploration.Publishing results in scientific journals and presenting at conferences.With a strong foundation in numerical modeling, machine learning, and data analysis, I bring expertise in programming (Python) and experience working with deep learning frameworks, especially in foundation models.

Sales Engineer | Driving Industry 4.0 InnovationsAs a Sales Engineer at Allmaster, I specialize in delivering cutting-edge machinery, components, and solutions aligned with Industry 4.0 principles. With expertise in artificial intelligence, mechanical engineering, and project management, I focus on automation and smart manufacturing to enhance operational efficiency.Key Contributions: • Project Oversight: Managed on-site assembly of units and ensured compliance with technical specifications. • Commissioning Leadership: Played a pivotal role in commissioning advanced systems, including providing technical training to optimize automation features. • Global Collaboration: Represented the company internationally, strengthening partnerships and driving innovation through technical training and trade events in Germany.

1. Wafer Map Defect and Pattern Recognition - Anomaly Detection in parametric level of CP tests in semiconductor manufacturing.Apply SOTA deep learning model such as Vision Transformers ( DeiT, BeiT, MAE) as the backbone network, combined with machine learning techniques, e.g., DBSACN, Isolation Forest, that robustly inference and find anomalous samples among +2,000 wafers, +1,000,000 parametric maps, across +20 IC products a day ranging from Power Management IC to mobile phone / TV IC.2. ETL pipeline from semiconductor fabrication and testing.Deploy Airflow task to orchestrate ETL task on GCP/GCS including BigQuery service and data clean-up, transformation, merging/joining the pandas dataframe/PyArrow parquet, that facilitates the inference of our downstream AI model.Compose DAGs and build the infra of Airflow to realize the automated ETL process, which is conduct daily to robustly yield +4,000,000 wafer maps a day.Manage and make use of GCP/GCS cloud resource where the source keys and raw data are stored.3. ML Application -Build EDA (Exploratory Data Analysis)Tool for data scientists with Streamlit.Democratize AI model by exploiting SOTA pre-trained model from CV, NLP problem. (Mainly withModularize full stack AI functionality and deploy services using Docker container and FastAPI.4. MLOps development Employed MLflow as a our MLOps tool to monitor the key metrics, the values of interested and hyper-parameters across deployed models and projects to efficiently keep track of the models and experiments.

1. NLP Task:Deploy BERT and its new variants as a service (RestAPI) to encode linguistic information for the data science team in the company.Develop BERT's downstream tasks and form analytic strategies, e.g., regression or classification by transforming and exploiting the unstructured linguistic data2. Anomaly Detection:Employ deep neural network (LSTM + Auto encoder) or ML algorithm (Isolation trees) + statistical modeling to catch the anomalies in quality checks of manufacturing processes that appears in form of time-dependent signals.It aims to improve the accuracy and high false negative/positive of the traditional testing.Meanwhile, the service pipeline provides an efficient and data-driven way to define the specs of quality testing of the company3. Time-Series forecasting (for Predictive maintenance):Apply recurrent neural networks (RNNs) based model and random processes to capture and predict the possible failures of manufacturing machines and actively provide suggestion for components replacement.Keyword:Machine learning, Deep learning, NLP, Time-series prediction, Probability modeling, Statistics, EDA (exploratory data analysis), Pattern recognition, Dimension reduction, Data visualization, SQL/NoSQL databaseApplied Python ML packages: NumPy, pandas, scikit-learn, hugging face transformers, XGboost, Tensorflow(Keras), PyTorchIn ML product development: Jupyter-notebook, Colab(GCP), FastAPI, streamlit, Docker container, Anaconda

Apr.2019-present - Chair of Engineering Risk Analysis, TUM• Developing scientific software for probabilistic modeling, risk analysis and uncertainty quantification( Python, MATLAB ) - Improved cross-entropy based importance sampling - Bayesian Inference Tools - PLS(Partial Least Square)-based PCE(Polynomial Chaos Expansion) algorithm for Machine Learning Surrogate- Translate MATLAB in-house codes to Python- Optimize Python codes of modeling of probability distribution and transformation- Introduce source control (git) to the team and help maintain the repository• Keyword: Bayesian Inference, Markov Chain Monte Carlo, Subset Simulation, Importance Sampling, Risk Analysis, Polynomial Chaos Expansion, Partial Least Square, Dimension Reduction, Surrogate Model, Machine LearningApr.–Sep.2018 - Chair of Computation in Engineering, TUM• Establishing a benchmark test by a manufactured solution for non-linear and phase-changing simulation ( C++, In-house code )• Keyword: Finite Element Analysis, Finite Cell Analysis, Numerical Modeling and Simulation

Apr.2019-present - Chair of Hydromechanics, TUM• Tutor - Computational Fluid Dynamics ( Finite Difference based numerical fluid dynamics simulation in MATLAB )Oct.2018–Feb.2019 - Chair of Computational Mechanics, TUM • Tutor - Finite Element Modeling, Simulation, and Validation ( Finite Element Analysis with ANSYS )

Full-time internship - Sensor Engineer, SYNTEC Technology CO. LTD., Taiwan Hsinchu Science Park • Establishing an interface for signals between accelerometer, encoder and PC-based controller with RC circuit and OP-AMP