A.E.G.I.S.

 Idea

A.E.G.I.S. - Aquatic Early-warning Guidance & Intelligence System

The problem

Twenty-five percent of European drinking water never reaches a customer.  540 milion € in just Rome. 7 milion € in Wrocław. YEARLY! Read the reports like water city index https://www.arcadis.com/pl-pl/insights/perspectives/europe/poland/water-city-index-2025

By 2028, the European Commission will introduce a legal threshold for permissible water losses - and the 13,000 small and medium utilities across the EU have neither the budget nor the infrastructure to comply. They find leaks when a citizen reports a flooded basement, not when satellites already know. BUT THE REAL VALUE LIES IN INFRASTRUCTURE KNOWLEDGE. It is much more reliable source of information than satellite images... It provides broader context to the satellite image.

The solution

AEGIS fuses three sources of truth into one operational decision:

• Athena - the orbit. Sentinel-1 SAR and Sentinel-2 optical imagery on a 6-day revisit, with NDMI z-scores computed against a baseline. We watch vegetation moisture for the statistical fingerprint of an underground leak, working through clouds and at night.
• Demeter - the surface. WorldCover land cover at 10 m resolution tells us where our signal is strong (greenery) and where to dispatch crews faster instead (built-up areas). The system knows its own limits.
• Hades - the underworld. A digital twin of the customer's network: pipe age, material, diameter, operating pressure, repair history. The same data utilities already have in their GIS - finally connected to something that can act on it.

These signals converge into a single number: the AEGIS Score (0–100), a dispatch priority for every pipe in the network. Not "wet patch here" - but "high-confidence leak on a 1973 cast-iron trunk main, dispatch the crew with the right cutter, before the basement floods on Friday."

What makes us different

We are not a prediction system. We are a catch-the-past system. We provide additional layer. We have validated AEGIS against 167 real water main failures from MPWiK Wrocław's 2025 incident database - and shown that the satellite signal was already shouting weeks before each failure was reported. EVERY STARTUP can use Satellite data. No competition can learn infrastructure as rapid as we can. The Zaporoska 24 February 2025 incident is our reference case: NDMI z-score reached -1.61 days before the rupture. The signal was loud. No one was listening. Today technicians scan about a kilometer circle. They are months too late. We provide much more precise information. We are listening.

EU space technologies

EU space technologies we used (and plan to extend during pilot/scale-up phase).

1) Copernicus Sentinel‑2 (multispectral optical EO) Data / signal: VIS-NIR-SWIR bands → vegetation & moisture indices (NDMI/NDVI/NDWI) and their change over time (T-1 / T0 / T+1). How we use it: we compute NDMI z‑scores against a 2018–2024 baseline (per‑month μ/σ) to detect anomalies over the water network and around recorded leaks. Value: provides an interpretable EO “fingerprint” of leaks in green areas (parks/vegetation) and supports early detection or post‑event confirmation. 

2) Copernicus Sentinel‑1 (SAR, C‑band) Data / signal: VV/VH backscatter (GRD) and (planned) InSAR coherence from SLC. How we use it / intend to use it: SAR adds an independent channel to confirm moisture/ground disturbance signals even when optical data is blocked by clouds. Value: works day/night and through clouds, improving robustness and reducing false alarms from purely optical indicators. 

3) Copernicus Land Monitoring / ESA WorldCover (10 m land cover) Data / signal: land‑cover classes (built‑up vs green) at 10 m. How we use it: we compute % built‑up and % green within a 200 m buffer around candidate points/segments. Value: a strong context filter (“is EO meaningful here?”) that rejects dense urban pixels and prioritizes locations where EO signals are verifiable (our DoD greenery criterion). 

4) Copernicus Data Space Ecosystem (CDSE) Technology: the European access layer for Copernicus data (catalog/APIs) used to obtain S1/S2 products and derived stacks. Value: enables scalable, repeatable, license‑free EO ingestion for production deployments without vendor lock‑in. 

5) (Planned) EUMETSAT H‑SAF (Hydrology SAF) Data / signal: operational satellite products for soil moisture and precipitation. Value: strengthens the “leak vs rainfall” separation and adds a hydrometeorological control signal to the fusion. 

6) (Planned) Copernicus DEM Data / signal: elevation and derivatives (slope, wetness index, flow paths). Value: helps explain where water would flow/accumulate after a leak, narrowing inspection areas and improving decision interpretability.

EU Space for Water we solve

CHALLENGE #1: Securing equitable and efficient access to water 

Meet the Team

Piotr Ryszko - Complex Networks Researcher, AI Architect of A.E.G.I.S, TU Darmstadt & European Space Agency Scholarship recipient, LiDAR Applications, GIS and Geospatial data expert.  
https://www.linkedin.com/in/piotr-ryszko/

Dominik Czech - Urbanization AI Researcher and Business Architect of A.E.G.I.S, coordinates talks with mentors and knowledge transfer. Software Engineer @ Nokia
https://www.linkedin.com/in/docz03 

Adrian Matusiak - Product Owner of A.E.G.I.S, PhD candidate @ Wrocław University of Science and Technology & Massachusetts Institute of Technology MISTI program joint researcher, Senior Software Engineer @ Asseco Poland, previously Desktop App Team Lead in Venture Builder Nanores.
https://www.linkedin.com/in/adrian-matusiak/

Michał Furgała - Machine Learning Engineer, Solopreneur Developer @ tanszykoszyk.pl, multiple and successful hackathon participant incl. HackNation, Ensemble AI hackathons. Highly skilled in web development and AI agent usage.
https://www.linkedin.com/in/michal-furgala/

Mikołaj Machalski - Computer Vision Researcher, Gen AI Developer @ Accenture, prev. Anomaly Detection Analyst @ Wrocław University of Science and Technology. In A.E.G.I.S responsible of data preparation and refinement. Provides proper technical readiness level for our solution.
https://www.linkedin.com/in/mikmach/

Jan Szyndralewicz - Computer Vision Researcher, Software Engineer @ Nokia / President & Co-founder of KN Emognition, IEEE PerCom 2026 Best Demo RunnerUp Award winner. Gives most crucial input, refinements and final optimizations.
https://www.linkedin.com/in/jan-szyndlarewicz/

Jędrzej Kocięcki - Data Engineer, Data Scientist @ Nokia, Apache Spark and big data pipelines expert. He is indeed our spark, since he headstarted our work and pushed thousands lines of code. Notebooks and reporting experts.
https://www.linkedin.com/in/j%C4%99drzej-koci%C4%99cki/

Mikołaj Olesiński - Data Analyst, DevOps/MLOPS, AI/ML Engineer @ Nokia
He transfers his knowledge in 5G technology to hydraulics and sewage systems.
https://www.linkedin.com/in/miko%C5%82aj-olesi%C5%84ski-376005221/


LINKS
Live demo: https://cassini-deploy.vercel.app/
Repository: https://github.com/00200200/cassini_hackathon