RP2040 and ESP32 AI
A two-board embedded AI demo that runs a 28.9-million-parameter language model directly on an ESP32-S3, with a separate RP2040 acting as a serial OLED display node for the generated text.
This project runs a small PLE TinyLM language model on an ESP32-S3, which performs the actual inference, while a separate RP2040 receives the generated text over serial and renders it on a 128x64 I2C OLED display. The model is stored in a custom flash partition and memory-mapped at runtime, with output-head and scratch buffers staged in PSRAM.
The repository includes the ESP32-S3 inference firmware, host-side tools for exporting and verifying the quantized model, the RP2040 display-node firmware, a workflow diagram and demo images. An earlier attempt to run the model on the RP2040 alone was dropped because the board's 264 KB of SRAM and lack of PSRAM could not support it at usable quality, and no cloud API is used at any point.