// the find
MicroAIInc/MicroAI-Atom-ESP32
MicroAI™ is an AI engine that can operate on low power edge and endpoint devices. It can learn the pattern of any and all time series data and can be used to detect anomalies or abnormalities, make one step ahead predictions/forecasts, and calculate the remaining life of entities (whether it is industrial machinery, small devices or the like).
MicroAI Atom is a proprietary anomaly detection and time-series prediction engine distributed as a precompiled static library (.a) for the ESP32. It targets industrial IoT use cases like predictive maintenance and network traffic anomaly detection. You get sample C code that wires the library to sensors and MQTT, but the actual AI engine is a black box.
The two sample projects (APM and Security) cover meaningfully different use cases and include working FreeRTOS + ESP-IDF scaffolding. The SD card config pattern for algorithm parameters is practical for field-deployed devices where you can't reflash to tune. Security channel monitoring running on an MCU without phoning home to a cloud inference endpoint is genuinely useful for air-gapped environments.
The core AI engine ships as a precompiled .a blob with no source — you cannot inspect, modify, or understand what the algorithm is actually doing. Last commit was December 2021 and it targets ESP-IDF v3/v4 patterns; the current ESP-IDF (v5.x) broke several APIs. The license is an evaluation agreement, not open source — you cannot use this commercially or in production without a separate deal with MicroAI. Documentation lives in PDFs inside the repo, which is a red flag for long-term maintenance.