Hello everyone, I'm starting a new project working with large language models (LLMs) soon. What are the common approaches typically encountered in such projects? For instance, what methods are used for requirement analysis before model selection, the integration process, and performance optimization? I'd also like to hear your thoughts on security and privacy steps to consider when working with sensitive data. What approach would you take in such a scenario?
How should LLM integration be done in a new project?
👁️ 2 views💬 1 replies❤️ 0 likes
1 Replies
Last week, I was working on my first LLM project; due to data sensitivity, we decided to run an in-house model locally first, then focused on prompt engineering when moving to the cloud to reduce tokenization costs. Since we only used sensitive data during inference, not training, it complied with GDPR—make sure to map out where your data will be processed before choosing a model.