Freelancer profile translated to English.
Back to original languageAbout Steffen
I help companies, data teams, and startups transition machine learning prototypes from **notebooks into stable, production-ready ML systems**.
Many ML projects work in experimentation but are difficult to integrate into daily operations: manual training runs, unclear model versions, lack of monitoring, difficult deployments, and little transparency about whether a model is still working reliably in production. This is exactly where I come in.
I develop MLOps solutions that enable **models to be trained, versioned, deployed, and monitored reproducibly**: ML APIs, automated training and retraining pipelines, experiment tracking, model registry workflows, drift and performance monitoring, CI/CD, and cloud deployments.
What makes me special: I combine data science with software engineering and cloud infrastructure. My focus is not only on the model itself but on the entire operational ML lifecycle.
I consciously use my portfolio projects on Customer Churn Prediction and Sales Forecasting as example applications for different ML problem types: classification, regression, and time series forecasting. The underlying MLOps architecture can be applied to many use cases, such as risk scoring, demand forecasting, lead scoring, inventory planning, or operational decision support.
Typical project types:
- Transitioning notebook prototypes into ML services
- Building prediction APIs with FastAPI
- MLflow experiment tracking & model registry
- Training and retraining pipelines with Prefect
- Dockerizing and cloud deployment of ML applications
- CI/CD for ML projects with GitHub Actions
- Monitoring model quality, drift, latency, and API metrics
- Infrastructure as Code with Terraform
- Deployment on Google Cloud Run
My goal: ML systems that not only work in a notebook but can be reliably, understandably, and expandably deployed in production.
German
Native or bilingual
English
Conversational
Remote only
Primarily works remotely
Experience
- FreelanceMLOps Blueprint for Classification & Business DecisioningApril 2026 - May 2026 (1 month)Development of a production-ready end-to-end MLOps showcase for a classification and decisioning system.Customer Churn Prediction serves as an example use case for typical business ML problems such as risk scoring, prioritization, and next-best-action decisions. The focus is not only on the model prediction but on the operationalization of the entire ML lifecycle: deployment, monitoring, retraining, model versioning, and translation of predictions into concrete business actions.A FastAPI-based prediction API with single and batch inference, expected-value-based decision logic for retention actions, customer prioritization, explainability endpoints, MLflow experiment tracking and model registry, as well as champion/challenger promotion with API model reload were implemented.For operational ML deployment, training and retraining workflows with Prefect, data validation, feature engineering, dataset versioning, delayed label monitoring, model performance evaluation, feature drift detection, and automatic retraining triggers were implemented.Additionally, the project includes Prometheus/Grafana monitoring for API metrics, a Streamlit dashboard for model and business monitoring, a Docker-based local development stack, and Terraform and GitHub Actions-based deployment on GCP Cloud Run, including a CI/CD pipeline with tests, smoke tests, Docker build, and Trivy scan.The project demonstrates reusable MLOps patterns for production-ready classification systems: model serving, model registry, monitoring, retraining, CI/CD, cloud deployment, and business logic based on model predictions.Tech Stack: Python, FastAPI, MLflow, Prefect, Docker, Terraform, GitHub Actions, Prometheus, Grafana, GCP, scikit-learn, Pandas.The complete code, including README, Docker setup, CI/CD workflow, and screenshots, is available via my linked GitHub profile.
- FreelanceMLOps Blueprint for Forecasting & Time Series PredictionFebruary 2026 - April 2026 (2 months)Development of a production-ready end-to-end MLOps showcase for time series forecasting, forecast monitoring, and automated retraining.Demand forecasting using a store sales example serves as a use case for forecasting, regression, and planning scenarios. The focus is not only on the forecast but on the operationalization of the system: training, API serving, monitoring, drift detection, evaluation, retraining, and deployment.The focus is on use cases where forecasts influence operational or business decisions, e.g., demand planning, sales forecasting, inventory planning, or capacity planning.A FastAPI forecasting API, structured prediction logs, training and retraining pipelines with Prefect, MLflow experiment tracking and model registry, and dataset versioning were implemented.For operational deployment, time-dependent feature engineering, forecast state handling, drift detection, forecast performance monitoring, alerting for degradation, and automatic retraining triggers were implemented.Additionally, the project includes a Docker-based local development stack, CI/CD with GitHub Actions, and Terraform and GCP-based cloud infrastructure with deployment on Google Cloud Run.The project simulates typical production problems: data changes, forecast errors increase, performance degrades, and new actual values become available with a delay. This transforms a static forecasting model into a controllable and continuously improving ML system.Reusable MLOps patterns are demonstrated: model serving, model registry, forecast monitoring, drift detection, retraining, CI/CD, and cloud deployment.Tech Stack: Python, FastAPI, MLflow, Prefect, Docker, Terraform, GitHub Actions, GCP, scikit-learn, XGBoost, Pandas.The complete code, including README, demo simulation, deployment documentation, and screenshots, is available via my linked GitHub profile.
- Technische Hochschule IngolstadtResearch Associate - Data Analysis & Simulation ModelingJuly 2021 - December 2024 (3 years and 5 months)- Development of data-driven simulation models for complex technical systems- Building reproducible analysis workflows for processing, validation, and evaluation of real measurement data- Conducting Design of Experiments (DoE), parameter optimization, and sensitivity analyses- Validating model behavior against real data and documenting model assumptions and limitations- Using model results for decision support, system optimization, and scenario analysis- Working with large datasets and structured analysis pipelines in several research projectsFocus: data-driven modeling, reproducible analyses, model validation, optimization, and decision support.
Recommendations
Be the first to recommend Steffen
Help this freelancer shine by sharing your experience working together.
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- Data Science & AI BootcampLe Wagon2025- machine learning fundamentals and model development - data processing and feature engineering - building and deploying ML applications
- M.Sc. Energy EngineeringFriedrich-Alexander-Universität Erlangen-Nürnberg2018- strong foundation in analytical and quantitative problem solving - experience working with complex technical systems - exposure to data-driven analysis in engineering contexts