Physicist (PhD) | Data Scientist

Project Portfolio

In the following, you´ll find an overview of my data science projects working in industry. Then you´ll find some research projects on machine learning working in academia. For the matter of completeness I appended some side-hustle projects I implemented in my spare time just for fun, curiosity or my eagerness to expand my knowledge in the field of machine learning and optimization.

Data Science Projects

Natural Language Processing Project:

Use Case Project Goal my Role Skills & Tools
AI Data App - SEO Content Creation using ChatGPT chat_gpt To implement a web-app for the generation of SEO content using ChatGPT I implemented a web-app to generate search-engine optimized (SEO) content. Python (numpy, pandas, spacy, nltk, googlesearch, openai, docs, requests, re), Jupyter, VSCode, streamlit

Top of Page

Healthcare Project:

Use Case Project Goal my Role Skills & Tools
Cerebral Disease Detection brain_volumetry To identity pathological cerebral changes I developed explainable classification ml-models providing disease propensity scores based on labeled MRI volumetry datasets. In addition, I documented relevant scientific publications on this topic. Magnetic Resonance Imaging, Image Analysis, Machine Learning, Data Preparation, Data Visualization, Explainable AI (shapley values), Python (numpy, pandas, sklearn, matplotlib, shap), Jupyter, Colab, VSCode, Gitlab

Top of Page

Marketing Analytics Projects:

Use Case Project Goal my Role Skills & Tools
Customer Value Prediction customer_value To improve customer value estimation I consulted the specialist department on how to improve an existing customer value model applying state-of-the-art machine learning methods. I further supported the data engineering team to continuously deploy new model versions. Machine Learning, Data Preparation, Data Visualization, SPSS Modeler
Use Case Project Goal my Role Skills & Tools
Customer Churn Prediction churn To improve customer churn prediction I consulted the specialist department on how to improve the existing customer churn model applying state-of-the-art machine learning methods. I further supported the data engineering team to continuously deploy new model versions. Machine Learning, Data Preparation, Data Visualization, SPSS Modeler
Use Case Project Goal my Role Skills & Tools
Customer Segmentation customer_segmentation To improve customer segmentation I developed a new customer segmentation model with improved performance compared to the existing model. I further supported the data engineering team to continuously deploy new model versions. Machine Learning, Data Preparation, Data Visualization, SPSS Modeler
Use Case Project Goal my Role Skills & Tools
Conversion Rate Optimization abschlussquoten_optimierung To improve conversion rates I developed a classification model based on tabular data providing propensity scores for conversion that could be used for priorization purposes. In addition, I advised the specialist department on the benefits and pitfalls during use of the model. Machine Learning, Data Preparation, Data Visualization, Python (numpy, pandas, matplotlib, sklearn), Jupyter
Use Case Project Goal my Role Skills & Tools
Price Optimization price_sensitivity To optimize product prices I developed a classification model based on tabular data as a foundation for further product price optimization. In addition, I advised the specialist department on the benefits and pitfalls during use of the model. Machine Learning, Data Preparation, Data Visualization, Python (numpy, pandas, matplotlib, sklearn, spacy), Jupyter, SPSS modeler
Use Case Project Goal my Role Skills & Tools
Data Deduplication entity_res To remove duplicated rows from large datasets I developed a classification-clustering mixture model based on tabular data to detect duplicates of names and addresses. In addition, I supported the development team with building data-pipelines and the deployment of the model. Machine Learning, Data Preparation, Data Visualization, Python (numpy, pandas, matplotlib, sklearn, dedupe), Jupyter, VSCode, git, SQL
Use Case Project Goal my Role Skills & Tools
Purchase Analysis maverick_buying To identity incorrect order processes I developed a classification model from historical tabular data providing propensity scores for incorrect orders. I deployed and updated the model in SPSS modeler. In addition, I advised the specialist department on the benefits and pitfalls during use of the model. Machine Learning, Data Preparation, Data Visualization, Natural Language Processing (bag-of-words), Python (numpy, pandas, sklearn, matplotlib), R, Jupyter, RStudio, Spyder, SPSS Modeler

Top of Page

Finance Project:

Use Case Project Goal my Role Skills & Tools
Credit Default creditworthyness To predict upcoming payment defaults I developed a classification model based on tabular time-series data providing the propensity scores of upcoming payment defaults. In addition, I advised the specialist department on the benefits and pitfalls during use of the model. Machine Learning, Deep Learning, Data Preparation, Data Visualization, Python (numpy, pandas, matplotlib, sklearn, keras), Jupyter, VSCode

Top of Page

Time Series Forecasting Projects:

Use Case Project Goal my Role Skills & Tools
Predictive Maintenance pred_maintenance To predict machine-failure of wind turbines I developed a classification model based on weather and sensory time-series data providing propensity scores for imminent failure of wind turbines. In addition, I advised the specialist department on the benefits and pitfalls during use of the model. Machine Learning, Data Preparation, Data Visualization, SPSS Modeler
Use Case Project Goal my Role Skills & Tools
Price Forecasting price To predict electricity market prices I developed a regression model based on historical time-series data to estimate the electricity market price for the next day. In addition, I advised the specialist department on the benefits and pitfalls during use of the model. Machine Learning, Deep-Learning, Data Preparation, Data Visualization, Python (numpy, pandas, matplotlib, sklearn, keras), Jupyter, VSCode
Use Case Project Goal my Role Skills & Tools
Meter Forecasting meter_reading_forecast To forecast meter-readings I developed a regression model based on historical time-series data to forecast meter-readings. In addition, I advised the specialist department on the benefits and pitfalls during use of the model. Machine Learning, Deep-Learning, Data Preparation, Data Visualization, Python (numpy, pandas, matplotlib, sklearn, keras), Jupyter, VSCode

Top of Page

 

Research Projects on Machine Learning

Use Case Project Goal my Role Skills & Tools
MRI-based OEF Mapping qBOLD_ann To improving mapping of the OEF (=oxygen extraction fraction) with artificlial neural networks I developed an artificial neural network regression model to improve the quality of OEF maps, very important for assessing tissue vitality, tumor diagnosis or planing radio therapy. (More details are provided in: Domsch et al., Magnetic Resonance in Medicine, 79(2), pp.890-899, 2018) Magnetic Resonance Imaging, Machine Learning, Artificial Neural Networks, Matlab
MRI-based OEF Mapping qBOLD_regularization To improve mapping of the OEF (=oxygen extraction fraction) with regularized regression I developed a regularized regression model to improve the quality of OEF maps, very important for assessing tissue vitality, tumor diagnosis or planing radio therapy. Further, I advised other colleagues on the development of this data modeling strategy. (More details are provided in: Domsch et al., Proc. ESMRMB Congress, Edinburgh, UK, 32, p.36, 2015); S. Thomas, S. Hubertus, S. Domsch and L. Schad, Proc. Int. Soc. Magn. Reson. Med., Paris, France, 26, p.2093, 2018 Magnetic Resonance Imaging, Machine Learning, Artificial Neural Networks, Matlab
Use Case Project Goal my Role Skills & Tools
MRI-based Diffusion Imaging diff_ann To research on improving the mapping of diffusion parameters with artificlial neural networks I advised other colleagues on the development of an artificial neural network regression model to improve the quality of diffusion parameter maps, very important for assessing tissue vitality, tumor diagnosis or planing radio therapy. (More details are provided in: Domsch and Bertleff et al., NMR Biomed, 30(12), 2017) Magnetic Resonance Imaging, Machine Learning, Artificial Neural Networks, Matlab

Top of Page

 

Side-Hustle Projects

Below you find a brief overview of some sample projects in the field of machine learning, natural language processing and optimization which I implemented in my spare time just for fun, curiosity or my eagerness to expand my knowlede in the field of machine learning and optimization. Please check out my github and docker-hub for more interesting data science projects and ai data apps.

Top of Page

 

Operations Research:

Use Case Project Goal Implementation Skills & Tools
AI Data App - Supply-Chain Optimization supply_chain_analytics To implement a supply-chain optimization web-app I implemented a supply-chain app to minimize the production costs by optimizing the production site locations and the respective quantities produced. The supply-chain was modeled using mixed-integer linear programming (MILP). Based on MILP combined with Monte-Carlo simulations, distributions were calculated for the production costs, the slack and the shadow prices and the production sites and their respective quantities. The app is deployed on dockerhub and in the streamlit-community-cloud. (More details are provided in my github) Mixed-Integer Linear Programming, Monte-Carlo Simulation, Python (numpy, pandas, matplotlib, PuLP), Jupyter, VSCode, streamlit, docker

Top of Page

 

Finance:

Use Case Project Goal Implementation Skills & Tools
AI Data App - Asset-Portfolio Optimization portfolio_optimization To implement a portfolio optimization web-app I implemented a finance portfolio optimization app to maximize the expected return for a given risk using monte-carlo simulations, genetic algorithms and artificial swarm intelligence. The app is deployed on dockerhub. (More details are provided in my github) Monte-Carlo Simulation, Machine Learning, Genetic Algorithms, Artificial Swarm Intelligence, Python (numpy, pandas, matplotlib, sklearn, shap, geneticalgorithm, pyswarm), Jupyter, VSCode, streamlit, docker, dockerhub
Use Case Project Goal Implementation Skills & Tools
AI Data App - Insurance Bill Prediction bill_prediction To implement both a web-app and web-api to predict insurance bills I implemented a regression model to predict insurance bills based on user input. The app is deployed on dockerhub. (More details are provided in my github) Machine Learning, Python (numpy, pandas, matplotlib, sklearn, flask-api), Jupyter, VSCode, streamlit, docker, dockerhub
Use Case Project Goal Implementation Skills & Tools
AI Data App - Credit Default Prediction default_pred_app Image courtesy: kindpng.com To implement a credit default prediction web-app I implemented a classification model to predict credit default risks. The app is deployed on dockerhub. (More details are provided in my github) Machine Learning, Python (numpy, pandas, matplotlib, sklearn, shap), Jupyter, VSCode, streamlit, docker, dockerhub, azure app service

Top of Page

 

Natural Language Processing:

Use Case Project Goal Implementation Skills & Tools
NLP - Text Recognition sentiment Image courtesy: thedatascientist.com To implement a text recognizer in Azure I implemented a language recognition model, trained on raw text, using Azure cognitive services to identify language, extract keywords and entities and analyse sentiments from hotel reviews. (The code and more details are provided in the github repo) Natural Language Processing, Machine Learning, Data Preparation, Python (numpy, pandas, sklearn, matplotlib), VSCode, Jupyter Notebook, Azure Cognitive Services

Top of Page

 

Healthcare:

Use Case Project Goal Implementation Skills & Tools
AI Data App - COVID-19 Prediction default_pred_app Image courtesy: ndr.de To implement a web-app for predicting COVID-19 patients in ICU I implemented a time-series regression model on public covid datasets, provided by www.ourworldindata.org, to predict the number of COVID-19 patients in ICU (=intensive care units) for different countries worldwide in real-time. Then I developed a streamlit app. The app is deployed on dockerhub. (More details are provided in my github) Auto-ML, Python (numpy, pandas, matplotlib, pycaret), Jupyter, VSCode, streamlit, docker, dockerhub

Top of Page

 

Image Analysis:

Use Case Project Goal Implementation Skills & Tools
Object Detection object_detection To implement a fruit detector in Azure I implemented a fruit detection model, trained on image data, using Azure cognitive services. (The code and more details are provided in the github repo) Image Processing, Machine Learning, Data Preparation, Data Visualization, Python (numpy, pandas, sklearn, matplotlib, pil), VSCode, Jupyter, Azure Cognitive Services

Top of Page