• Team lead (before: Senior Developer) of an agile team of 10 developers practicing SCRUM 
  
• Definition, implementation and maintenance of a calculation & validation engine for time-series data including data processing, statistical estimations and forecasting 
• Development and maintenance of a common function library
• Migration of data & business processes from the legacy system to the new platform
• Development of unit tests for all components (engine, library & data / business processes)

Support product developers in R&D with design of experiments and optimization of chemical formulations using a custom developed software utilizing black-box optimization.

• Design of Chemical Experiments and Formulation Optimization via Black-Box Optimization
• Contributed to design and implementation of benchmarking framework to test algorithm candidates
• Conducted benchmarks on algorithms, documented findings and presented feedback to product team
• Contributed to code clean-up and major refactoring of codebase
• Wrote unit tests to increase test coverage

Optimization of finished good quality utilizing predictive modelling of production process and subsequent optimization of key production parameters.

• Product Quality Optimization via Machine Learning and Mathematical Optimization

• Lead data product assessment to evaluate the findings of a previous POC and presented results to stakeholders
• Conducted exploratory data analysis (EDA)
• Improved existing data pre-processing
• Trained predictive models and evaluated their performance
• Assessed and improved formulation of mathematical optimization problems
• Repeated EDA, model building and evaluation on new data (additional products)
• Contributed to setting the scope and estimating budget for the data product

Initiative to standardize the data science toolchain / platform within the team and to provide requirements and feedback for improving the data ecosystem of Henkel in the long-term.

• Researched machine learning operations approaches and conducted interviews with data scientists in Henkel to gather requirements
• Exchanged ideas and aligned options with colleagues from the data platform team
• Conducted an evaluation of Databricks and AzureML against the collected requirements
• Deployed and configured an experimental AzureML workspace
• Implemented a demo application with common machine learning tasks like data preprocessing, model training incl. tracking of experiments and versioned model storage
• Conducted training sessions for the team
• Contributed to the professionalization and automation of AzureML workspace deployments

Credit risk assessment of customers based on a machine learning model that

predicts the probability of default using internally available data. The model is used to complement external rating information for bad-debt provisioning according to IFRS9.

• Project Management (budget, time booking, security, GDPR and architecture processes)
• Requirement and stakeholder management
• Definition of required data sources in collaboration with BI Team
• Request provisioning of data sources from BI Team including necessary change requests
• Implementation of the feature extraction pipeline including unit and integration tests in Spark and pytest on Databricks
• Design and implementation of the machine learning model including probability calibration
• Detailed model performance analysis
• Design and implementation of machine learning operations architecture on Azure in close collaboration with the DevOps Team. This included implementation of CI / CD pipelines, scheduling and orchestration of machine learning pipeline using brigade, automated model training on Kubernetes, performance tracking & model versioning in AzureML, logging & alerting using Azure ApplicationInsights and Azure Monitor.
• Detailed model performance analysis and consultation for expected impact analysis
• Presentation of concept to external auditors including technical correspondence
• Hypercare support after go-live
• In charge of maintenance and operations until end of 2021
• Presented two live-streamed upskilling sessions about machine learning & AI on the example of this project to an audience of hundreds of internal employees

Credit risk assessment of customers based on a machine learning model that predicts the probability of default using internally available data. The model is intended to replace an existing statistical solution to complement external rating information for bad-debt provisioning according to IFRS9.

• Project Management (budget, time booking, security, GDPR and architecture processes)
• Requirement and stakeholder management
• Defining and requesting data extracts for POC from BI Team
• Data exploration
• Data modelling and feature engineering of transactional data
• Training of machine learning model on unbalanced classes and performance evaluation
• Communication and presentation of proof-of-concept results to key stakeholders
• Presentation of concept to external auditors including technical correspondence

Energy demand forecasting for optimizing energy procurement. Future energy

demands of a production line was forecasted using predictive models based on the production schedule and sensor measurements.

• Requirement and stakeholder management
• Data exploration and feature engineering
• Trained time-series machine learning models (using scikit-learn, XGBoost and Prophet) and evaluated their performance
• Created an interactive dashboard using Plotly Dash which presented insights from the data exploration and visualized the model performance

The SalonLab Smart Analyzer is a handheld device that scans the hair structure through near infrared measurements. Hairdressers use the corresponding SalonLab App in their consultation and analysis process, in which the analysis of the infrared measurements is based on machine learning models to assess hair health.

• Collaborated in a high-impact projects with multiple external partners and internal teams
• Consulted on design and development of machine learning models
• Stress tested a commercial deployment solution for the model prediction microservice
• Collaborated with a cloud engineer to design an alternative deployment architecture that reduced prediction latencies from seconds to milliseconds. The adoption of the proposed architecture resulted in six figure savings of license costs compared to the commercial deployment solution.
• Implementation of the model prediction microservice that exposes a REST API developed in Python including logging and error reporting
• Performed minor changes and upgrades of the API in Dec 2019, Feb 2020, and Jul 2020

Optimization of collection efforts by assigning dunning strategies on customer level using a trained classifier.

• Data extraction from MS SQL source database and data modelling
• Data exploration
• Feature engineering to capture customer payment behavior
• Training of machine learning model on unbalanced classes and performance evaluation
• Presentation of proof-of-concept results to key stakeholders
• Mentored colleagues during implementation and roll-out of the product in 2019 / 2020

Data analysis and machine learning platform targeted at the chemical production and healthcare industry providing customizable modules for visualization, data exploration, assisted machine learning modelling, and anomaly detection in an intuitive manner to non-data professionals.

• Designed and developed a new anomaly detection module for the platform in close collaboration with a UI / UX designer and web developers
• Developed new capabilities and improved the existing capabilities of the platform?s machine learning backend using Python and MongoDB
• Contributed to custom module development for customers including translation and parallelization of existing code performing data processing and statistical modelling
• Contributed to transforming the platform deployment workflow towards using Docker containers and set up CI / CD pipelines with Jenkins
• Extended the unit- and integration-test suite to increase test coverage
• Contributed to data pre-processing pipelines for importing and connecting customer?s data sources to the platform

Consulting and development of sales forecasts for a customer in the agricultural machinery industry.

• Consulting the customer on sales forecasting
• Exploratory data analysis of sales data from multiple business units
• Development of several machine learning models (Prophet, XGBoost, Scikit Learn) for multiple business units based on historical sales data and external industry indices
• Performance analysis and comparison of the models
• Presentation of the results to the customer

Consulting, exploratory data analysis and anomaly detection on production monitoring data from an industrial process plant for a customer in the plant building industry.

• Consulting the customer on topics of anomaly detection and predictive maintenance
• Exploratory data analysis of production monitoring data
• Anomaly detection on production monitoring data
• Data pre-processing and data import to the Qlaym Qoactive Platform (see below)