OPERATING DOCUMENT / KUBEMETAL

MLOps Pipelines

Local LLM fine-tuning (LoRA/QLoRA), MLX quantization, and low-latency inference.

01 / Context02 / Procedure03 / Operations04 / Verification

Guide to authoring fine-tuning and inference pipelines in KubeMetal.

Supported Workflows

One-Click Model Downloads: HuggingFace and Ollama weight ingestion
LoRA / QLoRA Fine-Tuning: Supervised fine-tuning on local custom datasets
4-bit / 8-bit Quantization: 70% memory reduction via Apple MLX quantization
High-Speed Local Serving: OpenAI-compatible /v1/chat/completions endpoints

DOCUMENT TRACE

Return to the project record to see architecture, development pulse, proof and related field notes in one context.