Extensive experience in developing end-to-end data and AI solutions, building business cases, solution architecture, data pipelines, computing/storage management, (RESTful) APIs, technical requirements, full-stack solutions.
Extensive experience in developing Machine Learning/Deep Learning approaches such as CNNs, RNNs/LSTMs, and Transformers for diverse applications.
Led the project and managed all aspects, including reporting, coordination, and guiding a multidisciplinary development team to achieve milestones effectively.
Designed and developed a full-stack, interactive, and responsive web application and mobile app to facilitate access to urban data, ensuring usability and cross-platform compatibility.
Implemented an automated data processing workflow to integrate and transform heterogeneous datasets from over 10 APIs and data providers, enabling seamless data aggregation and standardization.
Deployed and hosted the web application on a cloud infrastructure, ensuring scalability, security, and high availability, and provided ongoing support and maintenance for smooth operations.
Technologies used: Backend: Python, Django; Frontend: React, Next.js, JavaScript; Database: PostgreSQL, PostGIS; Deployment: Docker, Nginx; Version Control & CI/CD: GitLab
Project Manager, Senior Machine Learning Engineer
Bundesinstitut für Bau-, Stadt- und Raumforschung (BBSR) / Bundesministeriums für Wohnen, Stadtentwicklung und Bauwesen (BMWSB)
Öffentlicher Dienst
1000-5000 Mitarbeiter
Led the project and managed all aspects, including reporting, coordination, and guiding a multidisciplinary engineering team to achieve milestones effectively.
Researched and developed advanced machine learning models, including deep neural networks (ResNet, Transformers) and classical approaches (SVM, Random Forest, XGBoost, LightGBM, CatBoost), for detecting vacant gaps using aerial imagery and geospatial data across Germany.
Designed and implemented data pipelines and workflows to preprocess data and automate the training and inference processes for vacant gap detection models, ensuring scalability and reproducibility.
Wrote publications, and conference presentations, and organized a workshop with over 20 participants, including representatives from municipalities, authorities, and research organizations.
Project Manager, Senior Machine Learning Engineer
DFKI GmbH
Internet und Informationstechnologie
1000-5000 Mitarbeiter
Co-Led the project, overseeing reporting, coordination, and guiding a multidisciplinary team to successfully achieve project milestones.
Conducted research and developed machine learning models (Transformers, LSTM, ResNet, XGBoost) for crop yield estimation using Sentinel-2 satellite data.
Designed and implemented a comprehensive machine learning framework enabling reproducible experiments for customers and project collaborators within the consortium.
Developed an inference pipeline for analyzing multi-source satellite imagery with a set of models developed within the project.
Automated the acquisition and preprocessing of Sentinel-2 imagery to create datasets for scaling models to new locations effectively.
Optimized machine learning models by refining data inputs (temporal dimensions, spatial context), network architecture, and performing hyperparameter optimization using Ray Tune.
Published research in conferences and delivered presentations to share project outcomes.
Project Manager, Senior Data Scientist
Landesamt für Umweltschutz Sachsen-Anhalt
Öffentlicher Dienst
250-500 Mitarbeiter
Led the project and managed all aspects, including reporting, coordination, and guiding a multidisciplinary development team to achieve milestones effectively.
Conducted quantitative and qualitative analyses of geospatial data to assess land use changes in relation to soil protection across the federal state of Sachsen-Anhalt.
Compiled analysis findings into a detailed report, incorporating results, tables, diagrams, and plots, which served as a basis for decision-making and publications by the LAU Sachsen-Anhalt.
Delivered presentations of project outcomes at the Ministerium für Wissenschaft, Energie, Klimaschutz und Umwelt (MWU) and at the LAU Sachsen-Anhalt colloquium.
Led the project, overseeing reporting, coordination, and guiding a multidisciplinary team to successfully achieve project milestones.
Conducted extensive research and development of advanced AI models, including Transformers, ResNet, LSTMs, and ConvLSTMs, to detect and analyze land use and land cover changes using high-cadence satellite imagery from sources such as Planet-Imagery and Sentinel-2. These models enabled tracking of temporal and spatial changes over the European surface.
Processed and analyzed over several terabytes of satellite data to compile one of the largest datasets for the Earth Observation research community. This dataset integrated Sentinel-2 and Planet-Imagery data, providing a valuable resource for advancing research in remote sensing and AI-driven geospatial analysis.
Designed and implemented an automated workflow to streamline the image analysis process, enabling efficient data preprocessing, model training, and inference. The workflow facilitated scalable and reproducible AI-driven analysis of large-scale satellite imagery data sources.
Developed an interactive web application to enable users to visualize and interact with both the satellite imagery and the analysis results. The application provided intuitive tools for exploring model outputs, comparing temporal changes, and gaining actionable insights for various use cases, such as environmental monitoring.
Led the project, overseeing reporting, coordination, and guiding a multidisciplinary team to successfully achieve project milestones.
Designed and developed a full-stack SAAS solution for the forestry domain to detect bark beetle infestations from drone imagery, integrating advanced machine learning workflows and geospatial processing.
Implemented an image analysis pipeline incorporating machine learning models for tree detection, health classification, and tree type identification using deep neural networks (e.g., ResNet). The pipeline included tasks such as orthophoto generation, keypoint detection, image alignment, inference with ML models, and exporting results into geoformats.
Built a user-centric web application using Django to visualize drone imagery and analysis results, supported by a PostgreSQL database and interactive visualizations with D3.js, ensuring seamless integration of data and insights.
Deployed, hosted, and provided ongoing support for the web application and analysis backend, leveraging Docker for scalability and reliability.
Conducted marketing initiatives, including creating a dedicated website for the solution and organizing several workshop to engage with potential customers and showcase the software’s capabilities.
Senior Machine Learning Engineer
MunichRe Services GmbH
Versicherungen
>10.000 Mitarbeiter
Integrated existing deep learning model designed for the detection of HVAC (Heating, Ventilation, and Air Conditioning) systems from aerial imagery, specifically focusing on urban areas. This enhancement involved refining the model's accuracy and efficiency, enabling it to better identify HVAC systems in complex urban environments, thereby improving the model's operational value for large-scale urban monitoring.
Implemented a fully automated processing workflow for handling aerial imagery. This workflow included geocoding, gathering, and preprocessing of aerial images, followed by running the images through the enhanced machine learning model for inference.
Developed a front-end interface for user interaction and a RESTful API to expose the entire image detection workflow as a service component. Deployed the solution using container technologies at the customer's cloud.
Project Manager, Senior Machine Learning Engineer
DFKI GmbH
Sonstiges
1000-5000 Mitarbeiter
Led the project and managed all aspects, including reporting, coordination, and guiding a multidisciplinary development team to achieve milestones effectively.
Collaborated with cross-functional teams and contributed to the development of a roadmap for a digital twin of Earth (DTE), leveraging Earth observation data and artificial intelligence to create an advanced, interactive model of the Earth's surface.
Developed and evaluated various generative neural networks for the harmonization of satellite imagery from different sensors (Landsat and Sentinel-2). The goal was to ensure compatibility of imagery from diverse sources, allowing existing models to seamlessly analyze and interpret the data on a unified, harmonized dataset. This process involved model adaptation and data acquisition, improving image quality, consistency, and aligning temporal and spatial resolutions, across multiple sensor types.
Showcasing one solution of the DTE, by applying generative models to obtain harmonized satellite imagery data and then quantify urban greenness worldwide. Used segmentation models such as U-Net to extract vegetation cover in urban areas. Additionally, implemented a user-friendly dashboard to visualize, explore, and inspect the results, allowing stakeholders to easily interact with the outcomes and gain insights on urban green spaces.
Project Manager, Machine Learning Engineer
European Space Agency (ESA)
Energie, Wasser und Umwelt
1000-5000 Mitarbeiter
Led the project, overseeing reporting, coordination, and guiding a multidisciplinary team to successfully achieve project milestones.
Conducted in-depth research and development of advanced deep neural networks, including ConvLSTMs, to analyze multitemporal satellite imagery for the classification of forest types. This work significantly enhanced the accuracy and timeliness of map product generation, offering a marked improvement over the current update cycles used by authorities. The project demonstrated the potential of AI-driven solutions to optimize workflows in forest monitoring and resource management.
Seamlessly integrated the machine learning model into the open-source framework eo-learn. This integration showcased the usage of the scalable framework and the model's capability to handle large-scale satellite imagery, enabling the efficient generation of high-quality map products while supporting reproducibility and ease of use within the Earth Observation community.
Organized and delivered engaging webinars to present the project's results to technical experts and domain specialists.
Project Manager, Machine Learning Engineer
Bundesministerium für Bildung und Forschung (BMBF)
Sonstiges
1000-5000 Mitarbeiter
Co-Led the project, overseeing reporting, coordination, and guiding a multidisciplinary team to successfully achieve project milestones.
Conducted extensive research on machine learning and deep learning models for drone image analysis, focusing on tree detection in forested areas. Explored various architectures and algorithms to enhance the accuracy and efficiency of tree detection from aerial imagery, contributing valuable insights to the field of remote sensing.
Contributed to the aqusition, preprocessing and development of a multimodal machine learning dataset by integrating satellite, aerial imagery, and forest inventory data across Germany. Collaborated with TUB to compile a new comprehensive dataset, which serves as a foundation for training and improving machine learning models in forestry domain.
Project Manager, Machine Learning Engineer
European Space Agency (ESA)
Sonstiges
1000-5000 Mitarbeiter
Development of advanced machine learning models to detect and quantify land use changes globally using satellite imagery. The models leveraged temporal and spatial data to identify patterns and changes.
Developed a comprehensive business plan, including market analysis, to identify potential customer segments and engagement strategies. Efforts included assessing market needs, exploring monetization opportunities, and initiating outreach to key stakeholders and potential clients to align the solution with user requirements.
Designed and implemented an interactive web application that allowed users to inspect areas of interest (AOIs) and analyze land use changes over time. The application provided dynamic visualizations and insights, enabling stakeholders to make informed decisions based on the data.
Project Manager, Machine Learning Engineer
FloodTags (WorldBank)
Sonstiges
5000-10.000 Mitarbeiter
Led the project and managed all aspects, including reporting, coordination, and guiding a multidisciplinary development team to achieve milestones effectively.
Developed an image analysis component for operational use, utilizing deep neural networks to analyze images in the context of flooding events. The image workflow was designed to automatically detect and assess flooding-related indicators such as flood presence, water levels, and flood damage. By training and leveraging advanced neural network architectures, the system provides reliable and real-time insights that are essential for flood monitoring, and response strategies.
Designed and implemented a RESTful API for seamless integration of the image analysis component into an existing cloud-based software solution.
Project Manager, Machine Learning Engineer
MunichRe
Versicherungen
>10.000 Mitarbeiter
Co-Led the project and managed all aspects, including reporting, coordination, and guiding a multidisciplinary development team to achieve milestones effectively.
Designed and implemented an image analysis workflow to estimate socio-economic properties using aerial imagery. This automated workflow leveraged geospatial data from different sources and advanced image processing techniques.
Conducted research and development on deep learning models for building instance segmentation, focusing on refining segmentation accuracy. Explored and evaluated various model architectures, loss functions, and pre- and post-processing strategies to handle diverse urban environments and building layouts effectively.
Developed an interactive dashboard to visualize segmentation results, socio-economic classifications, and the underlying imagery. The dashboard provided users with an intuitive interface to explore and analyze the data.
Project Manager, Machine Learning Engineer
MunichRe
Versicherungen
>10.000 Mitarbeiter
Co-Led the project and managed all aspects, including reporting, coordination, and guiding a multidisciplinary development team to achieve milestones effectively.
Designed and implemented an advanced image analysis workflow to extract building footprints from high-resolution aerial imagery. This workflow streamlined the process of detecting and analyzing urban structures via image segmentation models, enabling scalable and efficient data processing for large geographic areas.
Researched, developed, and evaluated a variety of deep learning models for building instance segmentation. This included experimenting with different model architectures, loss functions, and pre- and post-processing techniques to optimize segmentation accuracy and address challenges such as varying image resolutions, building shapes, and occlusions. Documented the experiments and results in a comprehensive technical report.
Zertifikate
Certified Professional Scrum Master (PSM 1)
Scrum.org
2024
Certified Kubernetes Administrator (CKA)
The Linux Foundation
2024
Kontaktanfrage
Einloggen & anfragen.
Das Kontaktformular ist nur für eingeloggte Nutzer verfügbar.