Train an LLM with me

I've got 12k USD compute credits left to spend. Who wants to train a small LLM with me? We've got a month. Open source publish, data can stay with us, research paper would be good.

2-7B size most likely would be possible. I'm thinking something task-specific. Generating ppt slides, focused on different language than English, generating svg icons, you name it.

Good fine-tuning project is fine as well.

Would appreciate someone who's up to date, reads hugging face news, saw some papers, knows what an Orca, Lora etc is. Good scraping skills welcome as well.

If you're not up for this and know good datasets and resources, throw them my way, thanks.