Private Markets, Angel and Secondary Investing, Sacra

Sacra

Bringing transparency to private markets so investors, founders, and employees can build on conviction.

Sacra delivers deeply researched insights on the private companies shaping tomorrow's markets. By surfacing the data, context, and operator perspective that are missing from private markets today, we help secondary buyers and sellers price equity accurately and unlock liquidity for the people building these businesses.

Private Markets, Angel and Secondary Investing

Bringing transparency to private markets so investors, founders, and employees can build on conviction. Sacra delivers deeply researched insights on the private companies shaping tomorrow's markets. By surfacing the data, context, and operator perspective that are missing from private markets today, we help secondary buyers and sellers price equity accurately and unlock liquidity for the people building these businesses.

Private Capital Markets

<ul><li>Companies like Ramp, Notion and Databricks take outputs from frontier models like OpenAI’s GPT-4 and use them to fine-tune smaller, cheaper and faster open source models like Mistral and Llama 3.1, reducing their costs by 90% for specific, repetitive tasks.</li><li>Fine-tuning platforms like OpenPipe ($6.7M raised), Predibase (Felicis, $28.5M raised) and Airtrain AI (Y Combinator, $3.2M raised) unbundle fine-tuning from the LLMs and rebundle it with MLOps, combining a software development kit (SDK) for collecting prompt-completion pairs with model training, eval, and deployment to launch and rapidly iterate on fine-tuned models.</li><li>We recognized that fine-tuning could greatly enhance capabilities and reliability, but it typically required having a specialist on staff. We realized that if we built the right workflow tool, we could replace that specialist and make it possible for anyone proficient in prompt engineering to successfully build a fine-tuned model.</li></ul>

Companies like Ramp, Notion and Databricks take outputs from frontier models like OpenAI’s GPT-4 and use them to fine-tune smaller, cheaper and faster open source models like Mistral and Llama 3.1, reducing their costs by 90% for specific, repetitive tasks.Fine-tuning platforms like OpenPipe ($6.7M raised), Predibase (Felicis, $28.5M raised) and Airtrain AI (Y Combinator, $3.2M raised) unbundle fine-tuning from the LLMs and rebundle it with MLOps, combining a software development kit (SDK) for collecting prompt-completion pairs with model training, eval, and deployment to launch and rapidly iterate on fine-tuned models.We recognized that fine-tuning could greatly enhance capabilities and reliability, but it typically required having a specialist on staff. We realized that if we built the right workflow tool, we could replace that specialist and make it possible for anyone proficient in prompt engineering to successfully build a fine-tuned model.

Kyle Corbitt, CEO Of OpenPipe, On The Future Of fine-tuning LLMs