Eleni Saka

Eleni Saka

Surveying and Geoinformatics Engineering - ML Researcher

Resume

The full resume can be found here

Education

National Technical University of Athens (NTUA)
School of Rural, Surveying and Geoinformatics Engineering (2020 – 2025)

Diploma Thesis: “Machine Learning for Wildfire Risk Prediction Using Positive–Unlabeled Data” — introduced a PU Learning approach for modelling wildfire ignition risk across the Mediterranean. Implemented in Python (PyTorch) on a 27-variable dataset combining meteorological parameters, vegetation indices (NDVI, LAI), and land-surface characteristics.
Thesis Supervisor: Dr. Ioannis Papoutsis, NTUA
Final grade: 9.41/10 | Thesis grade: 10/10

Work experience

Working on self-supervised learning and geospatial foundation models, with a focus on training and developing location encoders. Projects include extending the SatCLIP framework to more modalities and developing OSMGraphCLIP, a geospatial representation model that learns global location embeddings from OpenStreetMap data using contrastive learning. Currently exploring self-supervised learning and foundation model approaches applied to weather and climate data.

Developed scalable ML pipelines, distributed training frameworks, and privacy-preserving learning systems. Combined software engineering with research-driven experimentation, focused on efficient coding, automation, and reproducibility. Collaborated with an international research group; results and framework are being prepared for publication.

Participation in the CEMS research program. Execution of the project “Risk assessment of forest fires using machine learning, detailed mapping of burned areas and statistical analysis of historical fires” as part of the implementation of the research program titled “CEMS”.

During my internship at the Beyond Earth Observation Center of the National Observatory of Athens, I focused on wildfire mapping, risk assessment, and satellite data analysis using Sentinel-2 imagery, QGIS, and Python. I began with manual burned area mapping and statistical extraction, followed by the analysis of fire reports and active fire data from MODIS, VIIRS, and EFFIS. I contributed to the evaluation of the FireHub service using SEVIRI data and photointerpretation. As part of the Copernicus Emergency Management Service (EMS), I supported EMSN196 (Zambia) through land use change detection and data validation. I later worked on EMSN200 (Falakro) and EMSN197 (Azores), contributing to fire delineation, map production, geodatabase creation, and the development of Web Maps, Scenes, and Dashboards. In the final stages, I developed Python tools for automating the retrieval of cloud-free imagery and burned area mapping, and created a machine learning dataset to support fire risk forecasting.

Publications

Skills

Foreign Languages

Conferences and Seminars

Honours and awards