Bringing End-to-End AI/ML Model Lifecyle, all in one place.
eFabric™ is the "Unified ML Factory" for Edge AI. It is a low-code / no-code platform designed to build, train and deploy Artificial Intelligence models directly onto ultra-low-power chips, specifically Syntiant’s Neural Decision Processors (NDP).
Key Capabilities
eFabric consolidates the entire AI workflow of dataset management, training and flashing into one seamless workflow, eliminating the need for external tools.
It bridges the gap between data science and embedded engineering, allowing developers to deploy hardware-ready models in minutes without complex firmware coding.
Creates models specifically tuned for "always-on" battery-powered devices (microwatt scale) rather than general cloud AI.
eFabric eliminates deployment surprises by ensuring that components like AI Model, edge hardware used during test & validation are carried to the production environment.
A GUI-driven experience that allows software developers to deploy hardware-ready models without deep hardware knowledge.
Accelerate time-to-market. Single streamlined workflow from dataset import to flashing silicon. No need for complex toolchains and SDK coding.
Build models for power-efficient chips and memory-constrained embedded Linux MPUs.
Accelerate development by starting with an EVK PoC, then scale seamlessly with our production-ready SoM.
Fast-track your product launch by embedding our production-ready SoM to deliver autonomous edge intelligence.
A seamless journey from raw data to deployed silicon, consolidated into a single intuitive interface.
Import, organize, label datasets and manage in a Project.
Configure pre-processing and
feature extraction.
Choose an existing model architecture or design your own.
Live training metrics, logs,
and validation.
Export optimized model &
flash to chip.
Improve the models with refined dataset and other parameters.
Track, analyze, and optimize
Model performance.
Zero Friction Handover: From Cloud to Edge
Featured Product: TML120 and R2L100 Evaluation Kit.
eTML120 is a XENO+ Series Tiny ML Module which is a solderable module and is used for building Audio as well as Sensor data-based Edge AI enabled IoT devices
| Features | Description |
|---|---|
| MCU | ARM Cortex-M23 |
| Memory | 8MB NOR Flash |
| Audio-0 Interface | Audio port interface for external PDM Mics |
| Audio-1 Interface | PDM/I2S/TDM for external Audio input from Mic/audio |
| Audio-2 Interface | I2S/TDM for Audio output to be sent to Audio processor |
| DC Interface | DC interface for connecting any DC sensors |
| GPIO | 3x GPIOs from ML processor |
| Programming Interface | Serial UART-based programming interface with TML flashing tool |
| Host Interface | Serial UART-based interface to report ML classification events to host processor |
| Size | 15 x 15 mm |
| Operating Temp | -40 to 85 °C |
| Module Supply Voltage | 3V / 3.3V |
XENO+ Vision ML (Machine Learning) SOM (System-OnModule) module is a production ready solderable SOM Module and can be used as core CPU module for building Linux OS based Video/Image based edge AIML devices.
| Features | Description |
|---|---|
| Main CPU | Dual ARM Cortex A55 @ 1.2GHz |
| MCU | ARM Cortex M33 @ 200MHz |
| RAM | 16Bit 2GB DDR4-1600 |
| Flash | 8GB eMMC Flash (Up to 32GB) |
| Serial Flash | 16MB Serial NOR Boot flash |
| ISP | Simple ISP |
| AI Accelerator | DRP-AI Accelerator |
| Video CODEC | H.264 Enc/Dec 2K/30fps |
| OS | Linux OS |
| Features | Description |
|---|---|
| Graphics Engine | ARM Mali-G31 3D GPU |
| Camera Interface | 1× MIPI CSI-2 (4 lanes) |
| Display Interface | 1× MIPI DSI (4 lanes) |
| Ethernet | 2× RGMII Gigabit Ethernet interface |
| SD Interface | 1× SD interface for WiFi module or SD Card |
| USB | 2× USB2.0 for Camera or LTE module |
| Other | 4× I2C, 5× UART, 2× SPI, 8× ADC, 2× CAN-FD |
eFabric connects Seamlessly to the TML/R2L100 EVKs via a simple USB interface for instant model testing.
Flash the optimized model binary to the TML/R2L100 EVKs in seconds directly from the GUI.
Validate edge-AI performance on actual hardware with live confidence scores and logs.
Verify latency, power profiles, and accuracy in the real-world deployment environment.
Seamlessly transition validated models to high-volume manufacturing with optimized deployment toolsets.
Built for the Syntiant NDP and Renesas RZ/V Series edge AI Linux MPUs, for Audio/Vison/SLM
Rapidly expanding to support the
full spectrum of
ultra-low-power chips and other hardware, coming soon.
Perfect For
eFabric excels in Audio Classification, enabling developers to build robust models for complex auditory environments.
Identify and classify diverse environmental sounds and events.
Detect specific trigger words or phrases with high accuracy and low latency.
Classify background noise and suppress the noise for clear audio
Classify industrial machine sound patterns for operational health.
Unlock new application domains with support for sensor-based ML. eFabric supports complex temporal data patterns.
Accelerometer-based activity tracking and movement classification.
Precise hand-gesture recognition for touchless control interface.
Classify vibration patterns for anomaly detection in machinery.
RUL (Remaining Useful Life) & SOH (State of Health) prediction.
Perfect For
Perfect For
eFabric has evolved into a unified ML factory for all edge modalities. Next-wave enhancements enable advanced computer vision directly on edge chips.
Secure identity verification and access control on low-power devices.
Real-time occupancy tracking for smart buildings and retail.
Automated anomaly detection and vision-based threat monitoring.
One-platform for Audio, Sensor, and Vision model generation.
Battery health degradation monitoring, estimating RUL (Remaining Useful Life), Predicting State of Health. Thermal stress detection, Charging abnormality detection.
Glass break detection, Alarm detection, Baby cry detection, Animal sound recognition.
Wake word detection, Voice ID, Touchless control for gesture detection.
Vibration analysis, Acoustic fault detection, Anomaly detection.
Face recognition, People counting, Occupancy tracking, Smart surveillance.
