Adagos provides NeurEco, a physics‑driven AI platform that automatically creates compact, accurate digital twins from small sensor, simulation, or experimental datasets. By embedding physical laws into neural networks, it delivers parsimonious models for real‑time prediction, monitoring, and anomaly detection, reducing development time and computational cost for engineers in aerospace, automotive, energy, and manufacturing.
Funding
Funding not disclosed
Founders
Product
Problem
Engineers often need accurate predictive models of physical systems but face limited sensor, simulation, or experimental data, making model development costly, time‑consuming, and prone to over‑fitting. Traditional modeling approaches require extensive test campaigns and manual integration of physics, hindering rapid iteration across the V‑cycle.
Solution
Adagos offers NeurEco, a physics‑driven AI platform that automatically builds compact, parsimonious predictive digital twins from small datasets. By embedding physical principles into neural networks, NeurEco generates models that remain accurate over long‑term, real‑time predictions while requiring far fewer tests. The software supports data in CSV, NPY, and MAT formats and runs on Windows, Linux, and macOS, enabling seamless integration into existing engineering workflows. Engineers can use the generated models for monitoring, anomaly detection, and optimal operation, reducing development time, lowering costs, and improving system reliability throughout the V‑cycle.
Target Audience
Primary users are engineers and data scientists in industries such as aerospace, automotive, energy, and manufacturing who need predictive models for design, monitoring, and maintenance of complex physical systems.
Features
- Fully automatic creation of physics‑based neural networks optimized for the given dataset
- Parsimonious modeling approach that yields the smallest accurate model, minimizing computational load
- Support for sensor, simulation, or experimental data, even with limited sample sizes
- Fast training via progressive sample selection, dramatically reducing learning time
- Cross‑platform compatibility (Windows, Linux, macOS) and support for CSV, NPY, and MAT data formats
- Python API for integration into training, testing, and deployment pipelines, enabling in‑field model updates