We aim to use AI, particularly LLMs, to help humans learn more about the problematic world we humans are creating and participating in, and how we can reduce our negative impact on the planet soon, quickly and effectively.
Zap, our video interviewing tool with chat, needs to be got working again, ideally with Sveltekit and SurrealDB. This tool records video and audio in separate streams with the highest possible quality to make post production easier.
We need to create a secure and private web app to manage future interviews. This project will likely be based on SurrealDB and will require knowledge of SvelteKit.
Help us rebuild our energy charts creator. This project will involve a reactive interface (SvelteKit) and ideally use SurrealDB.
We want to create a local Large Language Model (LLM) of the highest quality possible within the limitations of using consumer grade GPUs. This project will ideally use QLoRA.
Help us fine-tune a local LLM on domain-specific knowledge. We may do some limited work with Claude 2, but its API pricing makes large-scale usage unfeasible. Main requirements will be some knowledge of AI and their APIs as well as good python skills.
Develop a book summarizer using a local LLM. This project will involve managing context window challenges which are currently on the cutting edge of local LLM development.
Help us create a voice-enhanced chat app. This app will allow users to speak to it and receive voice responses. One of its main purposes is to help people learn more about the (bad) state of the planet and how such knowledge could help make new and better decisions about how to address the worsening situation.
The app pulls energy charts from PDFs and puts them in a MySQL database. The chart area is often too large and may include extraneous text and/or other charts, which makes the indexing process very difficult and inaccurate. This needs improving and the search facility needs improving. The current version can be seen here: Chart Finder