Sw Ml Study

A visualization tool built by an ML study group to experiment with Mixture of Experts (MoE) and Engram concepts. The project explores model optimization techniques including agent optimization, recursive models, hierarchical structures, and KV cache compression.

Designed for researchers working with local LLMs on consumer GPUs (12GB-24GB VRAM). Developed using AI coding agents and cloud-based tools, with a focus on making larger MoE models accessible on smaller hardware.