Embedded-first C++26/C23 neural inference: float32+int8 on MCU/MPU/CPU. CMSIS-NN · ESP-NN · NMSIS-NN · XNNPACK. Peer-benched vs TFLM, microTVM, TF Lite, ONNX Runtime (Arm · Espressif · RISC-V · Apple).

Edge AI KWS on esp32s3 with DS-CNN model, TFlite Micro, ESP-NN SIMD, MFCC preprocessing

ESP32-S3 on-device char-LSTM inference engine: 51.5M int8 MACs/s (4.6x more efficient than float32 ESP-DSP), 32.59 chars/s with 1.6M param model. ESP-NN SIMD + SRAM tiling + dual-core. 8 no_std Rust crates on crates.io.

A bare-metal TinyML autonomous RC car running an INT8 MobileNetV2 model and sensor fusion on an ESP32-S3, achieving 40–60ms real-time inference latency via FreeRTOS without relying on Linux or the cloud.

Fork of the esp-tflite-micro, TFLite for Espressif Chipsets