Local LLM Fine-Tuning & Coding Agent Harness

Python QLoRA MLX LangGraph LLMs Subagents

Project Details

AI Systems Engineering project optimizing a local coding agent harness with subagent-driven development, custom skills, guardrails, and fine-tuning Gemma models with QLoRA on Apple Silicon.

Local LLM Fine-Tuning & Coding Agent Harness

Table of Contents

List of all the sections in this post. Click to jump to that section.

Local LLM Fine-Tuning & Coding Agent Harness

Project Overview

An AI Systems Engineering initiative focusing on local LLM execution, custom agent harnesses, and fine-tuning open models for software engineering workflows on Apple Silicon.

Highlights & Engineering Details

  • Coding Agent Harness: Engineered a modular local agent harness incorporating subagent orchestration, tool usage, custom skill libraries, and execution guardrails to achieve robust autonomous coding.
  • QLoRA Fine-Tuning: Fine-tuned local Gemma language models on a curated dataset of 10,000 conversational and code-generation instruction pairs using QLoRA, achieving high-quality code completion within consumer hardware constraints.

Project Links