I'm 15 and started learning ML at 14. The thing that frustrated me most wasn't the models - it was data prep. Finding datasets, cleaning them, formatting them for training took longer than actually building anything. So I built Vesper. It's an MCP-native tool that lets AI agents autonomously search, download, clean, and export datasets. Instead of a developer doing this manually, the agent calls Vesper and gets production-ready data in minutes. You can try it via CLI in under 5 minutes: npx vesper-wizard@latest Would love feedback from ML engineers and anyone building agent workflows.