I was able to get an AOgmaNeo agent to play Atari Pong on a Teensy 4.1 microcontroller. I believe it is the most resource-efficient general-purpose reinforcement learner!

https://github.com/222464/TeensyAtariPlayingAgent