From signal to decision
Sondera turns inspection signals into decisions.
AI-driven signal processing, defect detection, and decision support for NDE professionals, starting with terahertz and expanding across modalities.
The problem
Inspection produces more data than decisions.
01
Inspection data is underused
Facilities record terabytes of scans, then act on a fraction of them. Most of the signal content that could inform maintenance and qualification decisions is never analyzed.
02
Analysis is manual and expert-bound
Interpreting raw signals depends on a small number of specialists. Their judgment does not scale across sites, shifts, or growing inspection volumes.
03
Decisions lack traceability
Accept and reject calls often live in spreadsheets and reports detached from the underlying data. Auditing why a part passed is harder than it should be.
What Sondera does
One pipeline from raw waveform to signed-off report.
AI signal processing
Denoising, deconvolution, and feature extraction tuned to each modality's physics. Raw waveforms become clean, comparable, decision-ready data.
Defect detection and characterization
Machine learning models locate indications, estimate size and depth, and classify defect types, with confidence measures attached to every call.
Decision support and reporting
Results map to acceptance criteria and produce traceable, audit-ready reports, so engineers sign off on evidence rather than intuition.
Modality roadmap
Built for one modality first. Architected for all of them.
THz-TDS
In development, pilot openTerahertz time-domain spectroscopy for composites: coatings, delaminations, and layer thickness.
Learn more
Ultrasonic
PlannedUT waveform analysis for welds and structural components, building on the same signal pipeline.
Learn more
Micro-CT
PlannedVolumetric defect detection and porosity analysis for additive and cast parts.
Learn more
Built on research
Physics first, then machine learning.
Sondera grows out of peer-reviewed research in terahertz nondestructive evaluation and physics-based electromagnetic modeling. Models are trained and validated against simulated and measured data, and every algorithm is grounded in the wave physics of the modality it serves. No black boxes where a measurement standard applies.
- Peer-reviewed THz NDE research behind the core methods
- FDTD and finite element simulation for training and validation data
- Physics-informed models, benchmarked against reference measurements
- Confidence and uncertainty reported with every detection
Join the THz pilot program.
We are onboarding a small group of pilot partners working with composite materials. Leave your email and we will get in touch, or tell us about your inspection problem on the contact page.
Prefer to write? Contact us about the pilot