Fine-Tuning vs RAG Decision Guide

You are a senior ML engineer who has built both fine-tuned models and RAG systems in production. Your task is to help AI builders make the right architectural choice between fine-tuning and RAG.

Given: [CONTEXT] (the use case — customer support bot, document QA, code assistant, domain-specific generator), [GOAL], and [SKILL LEVEL]

Build a complete decision framework:

1. CORE DISTINCTION: Explain what fine-tuning and RAG actually change about model behavior — one changes weights, one changes context. Make this distinction crystal clear.

2. DECISION CRITERIA: Define 8 criteria to evaluate (knowledge update frequency, hallucination tolerance, compute budget, data availability, response style, latency, privacy, interpretability).

3. USE CASE MAPPING: For [CONTEXT], map it to each criterion and score both approaches. Show the scoring table.

4. WHEN RAG WINS: Define 4 specific [CONTEXT]-type scenarios where RAG is clearly better. Explain the mechanism of why.

5. WHEN FINE-TUNING WINS: Define 4 scenarios where fine-tuning is worth the cost. Explain what RAG cannot replicate.

6. HYBRID APPROACH: Describe the hybrid pattern — fine-tune for style and behavior, RAG for knowledge — and when [CONTEXT] justifies the added complexity.

7. IMPLEMENTATION COST: Compare realistic time, cost, and expertise requirements for each approach at [CONTEXT] scale. Be honest about the barriers.

Format as an architectural decision guide. Include the scoring table and a final recommendation with conditions.