embedded-ml

Machine Learning Systems: Foundations, Scaling, Agentic AI, and Physical AI (Vols I–IV) • Harvard CS249r | https://mlsysbook.ai

TinyML & Edge AI: On-device inference, model quantization, embedded ML, ultra-low-power AI for microcontrollers and IoT devices.

MINERVA - Minimal Inference Engine for Robust, Verifiable, and Authenticated ML. Encrypted, integrity-verified neural network inference for MCUs down to ATmega328P.

Pure-Rust prompt-injection detector with 1.5MB embedded MLP classifier. 98.40% accuracy, p50 14ms CPU inference, bindings for Python/JS/Go. Apache-2.0/MIT alternative to Rebuff (archived) and Lakera Guard.

Ahead-of-time compiler that fits ML models on embedded devices with hard memory budgets.

AIoT motion‑state detection pipeline: a simulated ESP32 + MPU6050 sensor (Velxio) streams accelerometer data over MQTT to Node‑RED, a TinyML model trained in Edge Impulse classifies Idle/Move/Shake on the edge, and predictions plus Shake alarms are visualized live on a ThingsBoard dashboard.

50.7x latency-optimized heterogeneous Vision Transformer (DeiT) hardware accelerator built on the Xilinx ZCU104 FPGA using DPU + custom HLS IPs.

Python ML for training a custom on-device cry model (knowledge-distilled from YAMNet, INT8, deployed on ESP32-S3)

ESP32 camera that escalates from gentle reminders to airhorn if you slouch

Suite of edge-AI tools and a web dashboard that help schools cut water, food, and energy waste by detecting leaks, sorting compost, and visualizing consumption in one conclusive dashboard.

Fajar Lang (fj) — Systems programming language for embedded ML & OS development. Compiler-enforced safety with @kernel/@device/@safe contexts. Rust-based compiler with Cranelift/LLVM backends. Made in Indonesia.

ChatTLM: a 10.9M-parameter language model that runs on a TI-Nspire CX II CAS graphing calculator, with the code, data and measurements behind the paper "Letting the Tools Do the Math"

Static INT8 AOT compiler that turns fixed PyTorch MCU models into standalone C11

Generative horror adventure with a quantized transformer running locally in a Game Boy-compatible ROM.

Novel DBSCAN-based FPGA system for automatic modulation classification and NDA SNR estimation. O(n²)→O(n) complexity reduction. 71.7% lower power than state-of-the-art.

Hardware-aware face detection on Samsung GT-S7392 (ARM Cortex-A9)

Curated Edge AI resources for computer vision & audio: hardware, frameworks, benchmarks, literature, and communities (excluding mobile).

End-to-end TinyML pipeline: gesture recognition on Arduino Nano 33 BLE Sense — 1D CNN (97.6% acc, 26.9 KB INT8) + 5 ML baselines, BLE→WebSocket→web dashboard.

Notes and resources from Qualcomm On-device AI course, provided by DeepLearningAI

Deploy and manage ML models at the edge — OPC-UA integration, PLC connectivity, real-time inference on embedded hardware for sub-millisecond decisions