Hi @seif,
Welcome to Google Cloud Community!
Here’s the answer to your questions:
- If I use Gemini from a model garden with vertex ai, does that mean i have my own private instance of Gemini in my infrastructure separate from Google's public Gemini model?
Using Gemini from Model Garden on Vertex AI doesn't create a completely isolated, private instance. Your data interacts with Google's infrastructure, but Vertex AI provides security controls to enhance privacy. Data is encrypted, access is governed, and Google complies with standards. Crucially, Google states your data is not used to train their models (verify terms). Utilize data source access control and Vertex AI security controls to restrict access and protect data. While not completely isolated, these features significantly reduce privacy risks.
- What other options (From GCP provided service) do i have to achieve this?
- Deploy Open-Source LLMs: Run models like Llama 2 directly on Compute Engine/GKE for maximum control over infrastructure and data. Requires more expertise. Apply GCP's security controls (IAM, networking).
- Vertex AI Custom Training: Fine-tune a pre-trained model within Vertex AI using your data.
- Private Service Connect: Enables private connectivity to external LLM services, avoiding the public internet.
Consider the trade-offs between ease of use, cost, control, and security. Carefully review Google Cloud documentation, terms of service, and configuration options, focusing on data source access control and Vertex AI security controls.
Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.
LinkedIn
Twitter
