Fine-Tuning: SFT, LoRA/QLoRA, RLHF and DPO
What each fine-tuning technique actually changes in the model, what it costs, when it beats RAG, and how to evaluate whether it worked
A deep, standalone treatment of fine-tuning for AI engineering interviews: the mechanics of SFT, LoRA/QLoRA, RLHF and DPO; realistic data requirements and where the data comes from; the sharp line between behaviour problems (fine-tune) and knowledge problems (RAG); training and serving cost order-of-magnitude; how to evaluate a fine-tune against catastrophic forgetting and a prompted baseline; and a worked decision scenario for a formatting-adherence bug in a support bot.
Practice questions (9)
-
View →
Picking the Right Technique for Three Different Requests
Advanced · Free -
View →
Diagnosing a Fine-Tuned Model That Memorized the Wrong Things
Advanced -
View →
Serving Per-Customer Fine-Tunes Without Per-Customer Deployments
Advanced -
View →
Justifying RLHF vs. DPO to a Skeptical Staff Engineer
Advanced -
View →
Designing the Evaluation Gate Before a Fine-Tune Ships
Advanced