Vinod Iyengar

Pocket LLM - The fastest neural search for your documents

Memorize 1000s of pages PDFs & Documents to search through them. Powered by AI and LLMs. Trained on your laptop. Fully private. Fully Free.

Add a comment

Replies

Best
Vinod Iyengar
Folks, I'm the head of product at ThirdAI and we are excited to get this PocketLLM tool out here for you all to try. The main aim is to showcase that we can build LLMs on laptops or any commodity CPUs. With the exponential growth of digital documents in recent years, finding the right document has become a hassle for many of us. Existing document search solutions rely heavily on server-side processing, raising concerns about data privacy and security. Enter ThirdAI PocketLLM, a new application from ThirdAI that allows you to train a neural network model locally on your laptop or air-gapped device to memorize an entire PDF or a folder full of PDFs, offering personalized and secure neural search. ThirdAI PocketLLM's innovative technology allows you to bypass the cloud and ensure your sensitive information stays private. Our model is trained on CPUs, making it affordable and accessible to everyone. You can train a 50M to 1B parameter model in minutes, without any specialized hardware. We understand the importance of privacy, which is why the entire ThirdAI PocketLLM can be trained locally. Your data is kept on your device, and no one, not even ThirdAI, can access it. Our solution is designed to offer personalized search that is specific to your needs, without ever compromising your privacy. ThirdAI PocketLLM's search is incredibly fast, thanks to the power of neural networks. Our solution uses embeddings, allowing for similarity search and natural language queries. We can't wait for you to try out ThirdAI PocketLLM!
Donovan
The privacy part of this app is brilliant, good luck!
Vinod Iyengar
@don_dikaio_arketekt_ thanks for your kind works. The same backend engine can be used for enterprise large scale deep learning as well to train massive private LLMs on their existing commodity hardware.
Eric Giroux
This looks awesome. There are like a bajillion use cases for this. Looks like a nice, simple solution. I also like your simple promo vids. Question: will it read handwritten text on PDFs?
Vinod Iyengar
@eric_giroux Thanks for your kind words. Definitely a lot more to come here. Currently hand written text and images are not supported. But it's on the roadmap to bring in OCR capabilities soon.