Personal AI is on a mission to build an AI for each individual to augment human biological memory and Carbon is excited to help them achieve this mission.
At the core of every Personal AI model is the user’s unified long-term memory, this long-term memory is comprised of multi-faceted, multi-modal data of an individual user - whether that is conversations, documents, websites, or external knowledge the user has accumulated over time. The robustness of uploading existing data and the continuous syncing of external knowledge to the user’s long-term memory is therefore crucial for the users’ Personal AI model to be accurate, comprehensive and up-to-date.
The team at Personal AI evaluated building in-house solutions based on Langchain as well as other embedded product integrations. They faced several challenges that needed to be addressed. These challenges included scaling for sudden bursts of workloads as their customers from around the world uploaded large files at ad-hoc times. Additionally, the speed of uploading data to long-term memory was a concern. The team also had to accommodate a wide range of integration requests, from Google Drive and OneDrive to Gmail and various websites.
When Sharon, Personal AI’s CTO / Co-Founder, reached out to us for a potential collaboration, we were thrilled to collaborate with them. Personal AI’s mission is to build an AI for each individual to augment human biological memory. This mission greatly resonated with our team and we wanted to contribute to achieving it
We helped Personal AI in integrating Google Drive and OneDrive. Personal AI customers could securely grant access to their own files and folders in these third-party applications to improve their AI's memory using Carbon.
Below is a breakdown of the Carbon products being leveraged in the Personal AI integration:
Personal AI receives notifications via webhooks when customers link their Google Drive or OneDrive accounts, and when files are processed.
Personal AI uses our unified API library to access content for each synced file, such as raw files, plain text, and metadata. This data is then fed into Carbon's in-house model pipeline, which combines all the information to create a Unified Memory for customer AIs.
Sharon and her team have been fantastic to work with, and their feedback has enhanced the Carbon developer experience.
We are excited to expand our partnership with Personal AI and introduce more data connectors this month for customers using their Unified Memory. We aim to create deep integrations so that their customer AIs can access the rich trove of knowledge from each customer through Carbon.