Case Studies
Connect external data to LLMs, no matter the source.
"We've been building ingestion pipelines for other third-party apps ourselves for a long time. However, it's proven much harder to build and maintain those integrations than we thought. Carbon helped us build it in just a weekend, delivering seamless integration to users, and now we can focus on what we do with peace of mind."
Mai Quang Tuấn
Mindpal
CTO/Founder
Challenge
To make their personalized AI agents as capable and knowledgeable as possible, Mindpal needed a way to easily connect a user's disparate data sources and ingest that unstructured data to train the AI. This included personal documents, notes, emails, and more stored across cloud storage providers like Google Drive, Dropbox, OneDrive and productivity apps like Notion. Manually integrating with each service and processing the data would be a massive technical undertaking.
Solution
Mindpal turned to Carbon and its universal data retrieval engine to solve this challenge. With Carbon's 25+ pre-built data source connectors, Mindpal was able to quickly and seamlessly stream user data from any source into their AI agent training pipeline. Key capabilities Mindpal leveraged include:
Ingesting unstructured data from sources like Google Drive, Dropbox, OneDrive, Notion, and more with just an API call
Automatically syncing new and updated content from connected sources on a custom schedule
Having Carbon process, clean, chunk and vectorize the ingested data to optimize it for AI training
Built-in enterprise-grade security with encryption of credentials and content
The ability to easily scale as Mindpal's user base and data processing needs grew
By building on top of Carbon's retrieval engine, Mindpal was able to focus on their core product - developing highly personalized and capable AI agents - rather than spending cycles on the complex data integration and processing work.
Leveraging Carbon allowed Mindpal to accelerate development and get to market faster with their innovative AI agent product. Carbon enabled Mindpal to:
Offer users the ability to connect data from 25+ popular apps and services to personalize their AI agent
Process and train on terabytes of unstructured user data from disparate sources
Quickly expand their agent's capabilities as new data sources were added
Ensure a secure architecture with industry-standard encryption and SOC 2 compliance
Conclusion
Mindpal continues to build out its "virtual taskforce" of AI agents with Carbon as a foundational data integration layer. With frictionless access to user data from any source, Mindpal is well-positioned to deliver on its mission of empowering individuals with personalized AI assistance to augment the human mind.
