IoT meets AI: edge AI, TinyML, anomaly detection and AI agents that consume and act on MQTT data streams.
A hands-on tutorial that sets up a Google Coral Edge TPU on a Raspberry Pi, runs a quantized MobileNet classifier and publishes every prediction as JSON to an MQTT broker.
How to design an LLM agent that reads device state from MQTT and sends commands safely, using a state cache, schema-constrained tools, a policy layer, human approval and a complete audit trail.
How camera-based defect detection works on a production line, how to design the inspection station, and how to publish pass/fail results over MQTT so PLCs, dashboards and traceability systems all get the same answer.
A tutorial on defining strict JSON schemas for device actions, validating every LLM tool call in code, asking the model to repair invalid calls and publishing only verified commands over MQTT 5.
TinyML is machine learning on microcontrollers that run on milliwatts and kilobytes. This explainer covers the constraints, the frameworks, real use cases and why tiny devices publish compact inference results over MQTT.
Models deployed on edge devices degrade quietly. This guide shows how to publish inference metrics over MQTT, measure drift centrally and raise alerts when a field model needs attention.
An industrial AI copilot is only as good as the data it can see. This guide shows how to ground an operator assistant in live, ISA-95 structured MQTT topics so its answers are current, traceable and safe.
A practical guide to predictive maintenance machine learning: which condition data to collect over MQTT, how to engineer features and labels, which models to start with, and how to validate them honestly.
Three simple, explainable methods for finding outliers in sensor streams, with Python implementations you can run against a live MQTT temperature feed and guidance on when each one fails.
A practical use case for AI HVAC optimization: which building data to collect over MQTT, how to learn a zone's thermal response, how to compute optimal start times and setpoints, and how to apply recommendations safely.
Create an account, get your broker credentials and connect your first device — no servers to install, patch or scale.
