Edge AI Meets Cloud Intelligence: Evaluating Favoriot for Lightweight Telemetry and Rapid Visualisation

Edge AI is rapidly reshaping how intelligent systems are deployed across smart infrastructure environments. Rather than streaming continuous video or raw sensor feeds to the cloud, modern architectures increasingly rely on real-time inference at the edge, transmitting only meaningful events for monitoring, analytics, and action.

In a recent experiment, Mohd Salleh, a Telecommunication Engineer, transitioned his object detection workflow from ThingsBoard to Favoriot, a Malaysian IoT platform, to evaluate its performance in supporting lightweight telemetry and rapid visualisation within an Edge AI pipeline.

His evaluation focused on four core areas:

• Real-time object detection at the edge
• Event-driven telemetry, transmitting data only when objects are detected
• Live monitoring through the Favoriot dashboard
• Night performance validation using standard CCTV infrastructure

This approach reflects a practical and scalable architecture for AI + IoT deployments.

From Continuous Streaming to Event-Driven Intelligence

Traditional IoT systems often rely on continuous data transmission, resulting in high bandwidth utilisation and unnecessary cloud processing. In contrast, Mohd Salleh’s architecture adopts an event-driven model:

• AI inference is performed locally on the edge device
• Only detection events are transmitted to the cloud
• Visualisation and monitoring are handled at the platform layer

This significantly reduces network load while maintaining operational visibility.

Such an approach is particularly relevant for smart infrastructure deployments in Malaysia, where scalability, cost control, and performance efficiency are key considerations.

The Next Step: Combining AI at the Edge and AI at the Platform

While edge inference enables real-time responsiveness, the next step is to integrate edge intelligence with platform-level analytics.

A hybrid architecture can be structured as follows:

AI at the Edge

• Object detection and filtering
• Immediate alerts for critical events
• Latency-sensitive decision making

AI at the Platform (Favoriot)

• Cross-device pattern analysis
• Trend detection over time
• Abnormal behaviour identification
• Detection frequency modelling
• Device health and telemetry correlation

Edge AI ensures fast local action. Platform AI enables broader intelligence across time, geography, and multiple nodes.

This combination transforms a monitoring system into a learning system.

Enhancing Telemetry Context for Smarter Insights

To further strengthen the deployment, additional contextual data can be transmitted alongside detection events:

• Model confidence scores
• Lighting condition indicators
• Edge processing latency
• CPU utilisation and system health metrics
• Camera uptime and connectivity status

With richer telemetry, Favoriot dashboards can evolve from simple visualisation tools into operational intelligence centres. This enables anomaly detection, predictive maintenance, and automated retraining triggers based on environmental performance conditions, such as nighttime detection degradation.

Moving from Monitoring to Closed-Loop Action

A key opportunity for improvement lies in closing the loop between detection and response.

By leveraging Favoriot’s rule engine and alert mechanisms, the system can be enhanced to:

• Trigger notifications only when detection thresholds are exceeded
• Escalate alerts based on frequency patterns
• Automatically log high-confidence detection clusters for dataset refinement
• Generate reports for performance benchmarking

This shifts the system from reactive monitoring to intelligent orchestration.

Scaling Beyond a Single Node

To validate real-world viability, the next logical step is multi-node deployment:

• Deploy across multiple CCTV units
• Compare inference consistency
• Measure telemetry efficiency at scale
• Evaluate edge vs. platform processing balance

Such testing provides deeper insight into latency, reliability, and scalability, particularly in smart city and industrial infrastructure scenarios.

Strengthening Malaysia’s AIoT Ecosystem

Experiments like this demonstrate how local platforms can support practical AI + IoT integration. As Malaysia continues to invest in smart infrastructure and digital capabilities, the integration of edge computing with cloud intelligence will become increasingly important.

Favoriot’s support for event-driven telemetry, real-time dashboards, and scalable analytics makes it a strong candidate for hybrid Edge + Platform AI deployments.

Small experiments often lay the groundwork for larger deployments. When designed thoughtfully, they can evolve into scalable systems that support intelligent infrastructure across industries.

For engineers or researchers interested in evaluating AI-at-edge combined with AI-at-platform capabilities, trial access can be arranged to explore deeper integration and advanced analytics features within Favoriot.

The future of AIoT will not be built on edge or cloud alone, but on the intelligent collaboration between both.