Hover a component
SoC + accelerator
ARM + NPU · 26 TOPS
Edge processing with an NPU for local inference: the decision stays on site.
Too many sensors, too little context
Plant sensors produce continuous streams that saturate networks and reach the center already stale. Sending everything raw to the cloud costs bandwidth and delays decisions exactly when a fault is emerging. The analysis has to move on site, next to the machine.
IoT gateway
Aggregates thousands of points from heterogeneous industrial protocols and normalizes them into a single data model.
Edge AI
Models inferred locally on an accelerator, with no dependence on connectivity to the center.
Predictive analytics
Anomaly detection and remaining-useful-life estimation from vibration, temperature and current signatures.
Store-and-forward
Buffers data during network outages and syncs it in order once the link returns.
Connect
Wire field sensors and buses to the FLUX gateway close to the asset being monitored.
Model
Define tags and deploy the predictive models trained on MONOLITH with CORTEX.
Analyze
The edge AI evaluates signals in real time and produces local indicators and alarms.
Forward
Only relevant events and aggregates are pushed to PULSE and archived in VAULT.
FLUX is the in-situ system that combines an IoT gateway with edge AI: it collects field signals, normalizes them, runs predictive analytics on site and forwards only relevant events and aggregates to PULSE and VAULT. The architecture is layered, from multiprotocol acquisition to accelerator-based inference, with store-and-forward to operate even without a network.
DIN-rail in cabinet
DIN-rail gateway close to the monitored asset, wired to sensors and field buses, powered from the local panel (Basic tier).
Concentration rack
Rack unit that concentrates many field nodes for a plant section, with greater inference and buffer capacity (Pro tier).
Ruggedized in situ
Fanless IP65 chassis for direct installation on machinery in harsh environments, resistant to dust, humidity and vibration (Enterprise tier).
Verified boot
Secure boot and TPM 2.0 ensure the gateway only starts signed firmware and models, anchoring node identity to the Zeklar network.
Encrypted buffer
The store-and-forward buffer is encrypted at rest: data held during an outage stays protected even if the device is tampered with.
mTLS channels and RBAC
Forwarding to PULSE and VAULT runs over mTLS; identity and entitlements come from AEGIS with role-based access control and audit.
Certifications
ISO 27001-aligned management; hardware with CE marking and industrial EMC (EN 61000-6-2) for plant environments.
What is the difference between FLUX and CORE?
FLUX is the IoT gateway and edge AI that collects and analyzes field signals; CORE is the downstream control node. They often work together, with FLUX upstream.
What happens if the network drops?
FLUX keeps acquiring and analyzing locally and buffers the data; on reconnection it syncs it in order and losslessly to PULSE and VAULT.
How do predictive models reach the device?
They are trained on MONOLITH with CORTEX and pushed to FLUX as signed artifacts, executed locally on the AI accelerator.
What signals does it analyze for predictive maintenance?
Typically vibration, temperature and current, from which it derives anomaly detection and remaining-useful-life estimation.
Does it really reduce network load?
Yes: instead of sending raw streams to the center, it forwards only relevant events and aggregates, cutting bandwidth and decision delay.
The mark encodes the mechanism, not the sector: la catena di giunti: il braccio che si articola.