OntoCraft
Ondevice AI · AIoT · IoT

AI runs without the cloud.

ESP32 · Matter · Edge AI · Local LLM. Deliver the same AI value across air-gapped environments, low-power devices, and every industrial site that demands data sovereignty.

Edge · Sovereign · Always-on

01

Data Sovereignty

Sensor and control data never leaves the device. Operates identically in air-gapped and classified environments.

02

Low Latency · Low Power

Instant on-device inference — no cloud round-trip. Lightweight model optimization that runs even on ESP32-class MCUs.

03

Metaprotocol Integration

Matter · MQTT · BLE · Thread unified in a single abstraction layer. Disparate device protocols converge on a shared ontology.

04

Lower Operating Cost

LLM API call costs eliminated. After initial deployment, operating cost is at electricity-bill level.

4-Layer Stack

From device to data pipeline.

L1Device LayerPhysical devices · sensors · connectivityESP32 / S3Matter 1.5 (Wi-Fi · Thread · BLE)Temp/humidity · light · accelerometerL2Edge AI InferenceLightweight model inference on-deviceTinyML · TF Lite MicroNormal/anomaly classification · voice commandDirect MCU inference (tens of ms)L3Local LLMOffline LLM + RAGLM Studio · Ollama · vLLMHome Assistant automationInternal document GraphRAGL4AIoT Data PipelineStream → store → visualizeKafka (topic stream)InfluxDB (time-series)Grafana (real-time dashboard)

Technology Stack

From device to data pipeline.

All four layers are open-source. Free to self-host.

01

Device

  • ESP32 / ESP32-S3 — Wi-Fi · BLE integrated MCU
  • Matter 1.5 — Wi-Fi · Thread · Ethernet metaprotocol
  • Thread Border Router · Wi-Fi mesh
  • MQTT · CoAP · BLE GATT protocol abstraction
  • Sensors: temperature/humidity · light · accelerometer · microphone
02

Edge AI

  • TinyML · lightweight classification model (normal/anomaly)
  • TensorFlow Lite Micro · ONNX Runtime
  • Real-time sensor inference + actuator control
  • Ontology-based semantic mapping (sensor · event · action)
03

Local LLM

  • LM Studio · Ollama · vLLM local execution
  • RAG · internal document Q&A
  • GraphRAG · knowledge graph combined reasoning
  • Home Assistant · n8n workflow automation
04

AIoT Data Pipeline

  • Kafka · real-time sensor event streaming
  • InfluxDB · time-series data storage
  • Grafana · operational KPI dashboard
  • MQTT · REST · WebSocket multi-protocol

Use Cases

Ready-to-deploy scenarios for the field.

Case 01

Smart Factory Predictive Maintenance

Vibration & temperature sensors → Edge AI anomaly detection → instant alert + parts recommendation

Case 02

Smart Home Automation

Matter devices + local LLM natural language control → no cloud dependency

Case 03

Field Work Assistant AI

Offline LLM + internal manual RAG → worker Q&A in air-gapped environments

Case 04

Energy Monitoring

Power sensors → InfluxDB → abnormal usage detection + automatic savings scenarios

Education Partnership

Build it yourself at
FDE Academy.

From hands-on ESP32-based ondevice AI to Matter smart home platforms — 11 new courses to get your hands dirty.

We'll demo AI running live
on real field devices.