I Trained My Own 30M Parameter AI to Run on an ESP32
www.youtube.com
I trained my own AI language model from scratch and got it running offline on an ESP32! 🐱In this video I'll show you how I trained a Transformer language mo...

4 Oct 2026 #ESP32 #Embedded #AI I trained my own AI language model from scratch and got it running offline on an ESP32! 🐱

In this video I’ll show you how I trained a Transformer language model with 30.46 million parameters and brought it to an ESP32-S3 DevKit and the Seeed Studio XIAO ESP32S3.

This AI has limited capabilities and takes time to respond. It was never meant to compete with modern LLMs. This was just a personal experiment to learn more about how language models work and whether I could make it work at all :)

I hope you find some of it useful! ❤️

GitHub: https://github.com/mmm1712/Local_AI_e…

hardware and components: https://amzn.to/4B00LyJ

Chapters:

  • 0:00 Intro
  • 0:22 Transformer Architecture & Design
  • 0:58 Dataset & Tokenization
  • 1:38 Training the Model
  • 2:29 Memory Limitations & Quantization (INT4)
  • 3:42 Running Locally on ESP32-S3 DevKit
  • 4:00 Porting to XIAO ESP32S3 & microSD Workaround
  • 5:19 Alternative 1: Google Gemini API
  • 5:52 Alternative 2: Local Ollama Server
  • 6:23 Outro & GitHub Details

#ESP32 #MachineLearning #Embedded #AI