AI × DataData→ModelModel→Value
Machine Learning Engineer and Data Scientist turning real-world and real-time data into models that ship—classification, computer vision, and forecasting, plus the pipelines behind them.
Five years across telecommunications, logistics, healthcare, and the public sector, building ML solutions that hold up in production.
Sharper models. Cleaner data. Decisions you can trust.
About Me
I am a Machine Learning Engineer and Data Scientist with a background in mathematics (MSc, TU Berlin) and five years building production machine learning across telecommunications, logistics, IT, healthcare, and smart-city projects.
I work end to end: scalable ML pipelines on Azure Databricks and Spark, computer vision for drone and satellite imagery, churn and revenue forecasting, and the MLOps that keeps models monitored, versioned, and retraining. I am equally at home in the data layer—ETL, warehouses, and Power BI reporting people actually use.
I like problems where the model has to earn its place: real-time data, messy inputs, and stakeholders who need results they can understand and trust.
Where I help most
- →Teams with promising models stuck in notebooks that need a path to production
- →Computer-vision problems on image, drone, or satellite data
- →Forecasting and predictive analytics—churn, revenue, demand—wired into reporting
- →Organizations building dependable data foundations: pipelines, warehouses, and BI
From Research to Reliable AI
The path from a promising idea to a model your team can depend on—at the seam of AI and data
Machine Learning Engineering
From framing the problem to a model that holds up on real data—classification, computer vision, and forecasting, built with the maths in mind and deployment in sight.
- →Computer vision for drone and satellite imagery (U-Net, ResNet, Azure Vision)
- →Predictive models—churn, revenue, demand—with rigorous evaluation
- →Reproducible training and MLOps with MLflow, Databricks, and CI/CD
- →A clear route from prototype to production, not a one-off notebook
Data Engineering & Analytics
The foundation under every good model: pipelines, warehouses, and reporting your team can actually trust.
- →Scalable ETL and data lakes on Spark, Databricks, and SQL Server
- →Feature engineering that turns raw signals into model-ready data
- →Power BI dashboards that turn forecasts into decisions
- →Data models and warehouse processes built to last
What I do
Capabilities spanning applied machine learning and dependable data
Machine Learning Engineering
Designing, training, and tuning models for classification, anomaly detection, and prediction with TensorFlow, scikit-learn, and Keras.
Computer Vision
Image and geospatial analysis—object detection and segmentation on drone and satellite imagery with OpenCV and Azure Vision.
Predictive Analytics & Forecasting
Churn, revenue, and demand forecasting on historical and real-time data, wired into automated reporting.
Data Engineering & ETL
Scalable pipelines, data lakes, and warehouses with Spark, Databricks, SQL Server, and SSIS.
Analytics & BI
Power BI dashboards and self-service analytics that make models and KPIs actionable.
MLOps & Delivery
Reproducible, monitored ML with MLflow, Databricks, GitLab CI/CD, and automated testing.
Selected Work
Production machine learning across industries—from real-time pipelines to computer vision
Customer Churn Prediction
Built a scalable ML pipeline for early churn detection and forecasting on historical and real-time data, with an automated flow from CRM and transactional systems into a Spark data lake and an agent-based reinforcement-learning module to optimise customer interactions.
Key Outcomes:
- ✓Early churn detection from historical and real-time signals
- ✓Gradient Boosting and Random Forest models tuned in an MLOps workflow
- ✓Interactive Power BI environment for churn and risk forecasts
Technology Stack:
Drone Image Inventory Analysis
Developed and operated an ML workflow for automated inventory analysis from high-resolution drone images, served through a containerised REST API on a highly available cluster with monitoring, logging, and automated retraining.
Key Outcomes:
- ✓Fully automated image processing via a scalable REST API
- ✓Highly available pipeline with monitoring and automated retraining
- ✓Modular pre- and post-processing for image optimisation and calibration
Technology Stack:
Satellite Image Segmentation
Designed an ML pipeline for satellite image analysis, comparing U-Net, ResNet, and Azure AI Vision for segmentation accuracy and runtime, with a robust augmentation strategy and reproducible tracking in MLflow and Databricks.
Key Outcomes:
- ✓Significant gains in segmentation accuracy across model approaches
- ✓Reproducible pipeline with model registration and versioning
- ✓Cloud benchmarking and cross-validation with Azure AI Vision
Technology Stack:
Revenue Forecasting & BI
Developed ML-based revenue forecasting integrated into an automated Power BI reporting ecosystem—exploratory analysis of sales and time-series data, segment and variance analysis, and a streamlined data-warehouse process.
Key Outcomes:
- ✓Improved revenue prediction accuracy with a custom forecasting model
- ✓Advanced analytics embedded in Power BI dashboards
- ✓Faster, cleaner data-warehouse retrieval after reimplementation
Technology Stack:
Explainable AI for Emergency Response
Built an explainable AI decision and communication system for emergency calls—an LTC-RNN processing time-dependent communication and location data in real time, with SHAP/LIME explainability and reinforcement-learning refinement on a Spark streaming architecture.
Key Outcomes:
- ✓Automated, explainable resource assignment for emergency calls
- ✓Real-time processing with an LTC-RNN and RL refinement
- ✓Scalable streaming pipeline with CI/CD and automated validation
Technology Stack:
Liquid Neural Networks for Anomaly Detection
Researched and implemented Liquid Time-Constant Recurrent Neural Networks (LTC) for time-series anomaly detection—generating test data, benchmarking against RNN and LSTM, and running the ML infrastructure on Kubernetes with a GitLab CI pipeline.
Key Outcomes:
- ✓Anomaly detection on time-series data, including EKG normalisation
- ✓Systematic comparison of RNN, LSTM, and LTC models
- ✓MLOps on Kubernetes with continuous deployment
Technology Stack:
Let's Work Together
If you have data worth learning from or a model worth shipping, let's talk.
No deck required—bring the real problem and we'll figure out the path together.