I'm dealing with some sensitive data that I want to process using a Language Learning Model (LLM). After processing, I need to encrypt this data into a secure bucket, but I don't want to use the default Key Management Service (KMS) for encryption. Importantly, I need to ensure that the data remains secure and is decrypted on the client side using something like WebCrypto. Essentially, I'm looking for methods to handle this data safely without exposing it to any cloud infrastructure.
If anyone can share some ideas or point me to relevant documentation, that would be really appreciated!
2 Answers
You might want to look into Ollama for setting up a local LLM, but if that's not feasible for you, consider setting up a hybrid approach where you do the sensitive parts locally and then send the less sensitive data to the cloud.
One option is to run everything on-premise. It'll give you complete control over your data and infrastructure, but you might lose some of the convenience cloud offers.
I can't go that route; I need the benefits of the cloud for scalability and flexibility.
I’m actually unable to do it locally, so I’m looking for alternatives that work well in a cloud environment.