<aside> 💡

10M raised, early. Taked with Batuhan

Volt and asymmetric finance did seed

</aside>

Doing synthetic data.

Different people from all around the world generating a single dataset, people getting rewards for contribution

Have partnership with Prime intellect for Intellect 2 100B params

• for this they are mostly doing post-training
• multi-language
• CoT, maths stuff etc.

How to verify data is good?

• exact check verification (same seed, same T) -verify the same
• Self reflection (same node checks)

Have an SDK

• But will do custom pipelines if required

Focusing on finetuning and RAG style

Data

• Text only
  • conversational datasets
• Also supporting function calling