Stack layer 02 · Edge / Embedded
Real-time intelligence starts at the edge.
Most inspection AI lives in the cloud, hours away from the measurement. The edge layer moves computation next to the inspection hardware, where acquisition is deterministic, sensors are synchronized, and a model can act while the part is still on the scanner. This layer is what separates Sondera from companies that only analyze inspection data after acquisition.
Platform under development · engineering services available now
Why NDE needs more than post-processing
Four reasons the computation must move.
01
Determinism
Inspection quality depends on when samples are taken, not just how many. Acquisition, encoder counts, and multi-sensor triggers must land on a hard schedule that cloud round-trips cannot honor.
02
Throughput
A scanning system produces waveforms faster than most links can ship them raw. Local preprocessing and feature extraction reduce data at the source, keeping full physics where it is cheap to have it.
03
Synchronization
Multi-modal and scanned measurements only fuse correctly when sensing channels, encoders, and motion are aligned in time. Synchronization is an embedded problem, solved next to the hardware.
04
Autonomy
Adaptive and closed-loop inspection needs decisions in milliseconds, at the instrument. Edge inference is the prerequisite for everything the automation layer will do.
The embedded chain
From transducer to action, on the instrument.
01
Sensor
physical response
02
ADC / DAQ
deterministic acquisition
03
FPGA
high-throughput preprocessing
04
MCU / SoC
system control
05
RTOS
real-time scheduling
06
Signal processing
embedded DSP
07
Edge AI
local inference
08
Control / Comms
act and report
The same chain scales from a single benchtop instrument to a multi-sensor robotic cell. What changes is where each block runs, not what the architecture is.
Capabilities
What this layer covers.
Sondera offers embedded software engineering for NDE and sensing systems as a service today: data acquisition, embedded DSP, device communications, and hardware/software co-design for instruments and inspection cells. The labels here describe the Sondera platform's own edge capability, not what the engineering team can build for a client.
Our founder teaches computer architecture, digital systems, and embedded systems at the university level. That discipline is a core company skill, not an outsourced one.
- Data acquisition and sensor interfacesIn development
- Embedded signal processingIn development
- MCU and SoC architecturesIn development
- Hardware/software co-designIn development
- Sensor and encoder synchronizationIn development
- Device communications and high-speed transferIn development
- FPGA-based processingRoadmap
- Real-time operating systems and schedulingIn development
- Edge inferenceRoadmap
- Scanner and actuator interfacesRoadmap
Key message