When to fine-tune vs. prompt, LoRA, QLoRA, evaluation, and deployment.
Know when to fine-tune, how to do it without wasting a budget, and how to run models in production — from LoRA and QLoRA to serving, monitoring, and cost control.
ML/AI engineers past their first prompt-engineered feature, and the platform teams supporting them.
The guides are numbered — read in order for the curriculum path, or jump straight to the one you need. Each card is self-contained.
The decision framework for adapting an LLM.
Continuing training to shift a model's weights.
Quality, format, size and common pitfalls.
Parameter-efficient methods that cut cost.
Hands-on walkthrough on Llama/Mistral-class models.
How models are aligned to be helpful.
Metrics, benchmarks, avoiding overfitting.
What changes when LLMs go to production.
Hosting, APIs, throughput and reliability.
Watching production systems that can degrade silently.
Automated regression testing for LLM apps.
Input/output filtering and abuse prevention.
Practical levers to cut the token bill.