For a specific scenario of open source Large Language Model application, we can use SFT (e.g, LoRA, PEFT) to SFT the general LLM for a big improvement of effect on a specific domain problem.
e.g,:
1. review or seek a special term in a business contract.
2. knowledge finding , in a pile of documents, even multi-modal data types, .docx, .pdf, .png, .jpeg, .xslx, etc.
3. agentic architecture application design, and implementation.
4. other data-driven and Machine-Learning driven application dev.