finetuned-ai-rag (AI Engineering & RAG Knowledge Assistant)

Model Description

This model is a fine-tuned version of Microsoft's Phi-3-mini-4k-instruct. I fine-tuned this model specifically to act as a technical tutor and assistant for AI Engineering, RAG (Retrieval-Augmented Generation), and LLM development.

Training Data

The model was fine-tuned using a custom curated dataset of high-quality technical Q&A pairs focused on:

• Building RAG (Retrieval-Augmented Generation) systems
• Large Language Model (LLM) fundamentals
• Fine-tuning techniques and LoRA adapters
• AI Engineering career paths and best practices

Training Details

• Base Model: unsloth/Phi-3-mini-4k-instruct
• Framework: Unsloth, TRL (Supervised Fine-Tuning)
• Hardware: NVIDIA T4 GPU (Google Colab)
• Quantization: 4-bit quantization for memory efficiency
• Training Method: LoRA (Low-Rank Adaptation) with 4-bit base model

Use Cases

This model is designed for:

1. Educational purposes: Helping developers understand RAG and LLM concepts.
2. RAG Prototyping: Answering technical questions about retrieval-augmented generation.
3. Fine-tuning demonstrations: Showing the capabilities of efficient fine-tuning techniques.

Intended Use

This is a demonstration model showcasing my ability to fine-tune LLMs using modern libraries (Unsloth, TRL, PEFT). It is intended for portfolio proof-of-work and educational Q&A.