Since joining Æthos-Sundai, an ethical-AI-accelerator slash hacker-foundation-from-Harvard-and-MIT, I've been toying with open-source LLMs. Many projects, apps, etc, tend to be hosted publically, prompting and saving data somewhere in the cloud.

Since my first Sundai @ MIT IHQ, I decided to provide simple code for the community to use LLMs in a private manner.

This model uses a basic RAG approach: feeding a PDF knowledge base into an open-source AI, fine-tuning it, and prompting it to respond to questions with references to its sources. It could be useful for researchers or as a starting tutorial for RAG implementations.

If you have questions regarding open-source ML, feel free to contact me!

🥷🏼👾 douzog(@)proton.me