sabucha
...joined a month ago, and has 1 karma
submissions / comments / favourites
Building MindMorph Labs — experimenting with CPU/NPU-first training methods for large language models (no GPUs, no full backprop). Currently working on N3L, a forward-only, low-rank local learning approach.
Exploring optimization, edge compute, and deep-tech design philosophy. Curious to connect with anyone working on efficient training or distributed AI systems.