Research
Sondera is built on peer-reviewed research, and every research output maps onto the same four layers as the technology: Sensing, Edge / Embedded, Intelligence, Automation. No claim on this site reaches further than the work behind it.
Outputs
Composites Part B: Engineering
Deep learning applied to THz time-domain measurements of impact-damaged glass fiber composites: detection, localization, and severity quantification of the damage mode aerospace fears most.
NDT & E International
Volumetric micro-CT imaging as independent ground truth for THz-based damage assessment, closing the loop between what the model predicts and what the material actually contains.
Founding research program
Combining independent physical channels, THz and volumetric imaging, into one damage assessment, the research basis for the stack's multi-modal intelligence.
US provisional patent application, 2026
Finite-element impact damage simulation coupled to full-wave measurement simulation, producing large, exactly labeled training datasets. Modality-independent by design: the same architecture serves THz, ultrasonic, and other inspection physics.
Doctoral dissertation, Georgia Institute of Technology
The founding body of work: measurement, simulation, and learning methods for THz nondestructive evaluation of composites, from which Sondera's intelligence layer is built.
Open directions
Sensing
THz response of composite structures across materials and damage modes; extension of the measurement architecture to guided waves.
Edge / Embedded
Real-time acquisition architectures, synchronized multi-sensor measurement, and embedded inference for inspection instruments.
Intelligence
Physics-informed defect detection, sim-to-real transfer, uncertainty quantification, and instrument-conditioned synthetic data.
Automation
Adaptive scanning policies, confidence-driven reinspection, and the control architecture for closed-loop NDE.
We collaborate with university groups and industrial labs on specimens, measurements, and validation studies, and we are glad to hear about datasets that deserve better analysis. Leave your email or write to us directly.