Build a Private AI Assistant on Cloud GPU

What you will build

A private, ChatGPT-style assistant that cites your own documents.

Starting point: you only need a Vultr account. This challenge is self-contained, so you can do it without having done any other.

The plan

1. Deploy a Cloud GPU instance and install drivers
2. Serve an open model with vLLM
3. Add a chat UI such as Open WebUI, behind TLS and login
4. Index at least 20 of your own documents
5. Write 10 test questions with known answers and measure how often retrieval finds the right passage
6. Improve retrieval and show what changed

vllm serve MODEL_NAME --max-model-len 8192

How you will be judged

• Endpoint is not open to the whole internet
• Answers cite source passages
• A measured hit rate, before and after your improvement

Stop or destroy the GPU instance when you are not using it. GPUs are billed while they exist.