LMGNU develops frameworks for local-first Large Language Model (LLM) systems. We design practical, optimized architectures that enable both training and inference directly on local infrastructure - prioritizing data privacy, low latency, and resource efficiency.
| Repository | Purpose |
|---|---|
| llm.cpp | Core inference engine optimized for localized, lightweight training. |
| LLM-Engine | python based resource management, pipeline orchestration. |
| prism | Tooling and interfaces optimized for local data processing. |
| nanollm | Compact LLM implementations simple and local training. |
