When you look at all of the levers you could pull to make your LLM app better, the number of possible configurations is infinite.
In this guide, we review some popular techniques that you might hear about while building with LLMs.
The thing to remember when considering any of these techniques is that they help your LLM products behave the way you want them to — by introducing more context, introducing more constraints, or both.
- Prompt Engineering
- Prompt Chaining
- Retrieval-Augmented Generation (RAG)
- Vector Databases & Embeddings
- Fine-Tuning with a Cloud API Model
- Instruction-Tuning with a Cloud API Model
- The API Model Frontier
- PEFT with an Open-Source Model
- Prompt-Tuning with an Open-Source Model
- Fine-Tuning with an Open-Source Model
- Pre-Training a Proprietary Foundation Model
- Foundation Model Orchestration (FOMO)