Behind the Craft Complete AI Course on Prompting, Evals, RAG, and Fine-Tuning | Adam Loving (Meta)
61 snips
Jun 1, 2025 
In this discussion, Adam Loving, an AI partner engineer at Meta, shares his expertise in integrating AI into products for numerous companies. He breaks down the essentials of crafting effective AI prompts and the nuances of evaluation strategies. Adam explains the advantages of retrieval-augmented generation over fine-tuning models, demystifies vector databases, and highlights the potential of open-source AI like Meta's Llama 4. His insights are invaluable for anyone looking to enhance AI functionality in their business.
AI Snips
Chapters
Books
Transcript
Episode notes
Granular AI Grading Methods
- Grade AI answers on elements like accuracy, tone, and hallucination.
- Use plus-one scoring per correct item to improve evaluation granularity.
Iterate to Focused AI Models
- Start with general AI features and then narrow focus with evals to specific tasks.
- Fine-tune models for high-volume, specialized tasks to improve reliability and cost-effectiveness.
Build Strong Human Eval Foundations
- Invest early in creating golden answer sets for human evals to ensure consistent quality.
- Edit AI-generated baseline answers to build better ground truth datasets.
