The context windows of AI models, which indicate the capacity of a model to “remember” information, have increased over time. However, researchers have suggested new ways to increase long -term memory of AI models, as they often cannot hold context on several sessions.
The 19 -year -old founder, Dhravya Shah, tries to solve problems in this area by creating a memory solution, called Supermemory, for AI applications.
Originally from Mumbai in India, Shah began to build robots and consumer -oriented applications a few years ago. He even sold his bot which formatted tweets in beautiful screenshots to the Hypufury social media tool.
The founder, who was preparing for an entrance exam to go to the IIT (Indian Institute of Technology), earned money from this sale and decided to move to the United States to attend Arizona State University instead.
After moving, he confused himself to build something new every week for 40 weeks. During one of these weeks, he built Supermemory (which was initially called any context) and put it on Github. At that time, the tool allowed you to chat with your Twitter bookmarks.
The current version of the tool extracts “memories” or information from unstructured data and helps applications to better understand the context.
Shah obtained an internship in Cloudflare in 2024, where he worked on AI and infrastructure. He then worked as responsible for relations with developers in the company. Meanwhile, advisers, including the CTO Dane Knecht from Cloudflare, asked him to transform Supermemory into a product.
This year, he decided to build Supermemory full -time.
Supermemory, now described as a universal memory API for AI applications, builds a data graphic based on the data he process and personalizes the context of users. For example, he can take care of the interrogation of one month’s entries for a writing or journalization application, or looking for a messaging application. As the solution allows multimodal inputs, it could also allow a video editor to recover the relevant assets of a library for a particular prompt.
The startup can ingest any type of data, says the company, including files, documents, cats, projects, emails, PDF and application data flows. Its chatbot and notifying function allows users to add memories to text, add files or links and connect to applications like Google Drive, OneDrive or the concept. There is also a chrome extension that allows you to easily add notes from a website.
“Our basic force is to extract information from any unstructured data type and give applications more context on users. While we are working on multimodal data, our solution is suitable for all kinds of AI applications going from messaging customers to video editors,” said Shah.
Supermemory obtained start -up of $ 2.6 million led by Susa Ventures, Browder Capital and SF1.VC. La Ronde also includes individual investors like Knecht from Cloudflare, the leader of Google Ai Jeff Dean, the product manager Deepmind Logan Kilpatrick, the founder of Sentry David Cramer and the leaders of Openai, Meta and Google.
Shah said that at one point, the combination also approached him to join one of his prizes, but he already had investors on board, so the moment did not work.
Joshua Browder, founder and CEO of the startup of “Robot Lawyer”, Donotpay, who directs Browder Capital as GP Solo, was impressed by the tenacity of Shah.
“I connected with Dhravya on X, and what struck me was the speed with which he moves and built things, which prompted me to invest in him,” he said.
The company has several existing customers, notably Cluelley, Ai Montra video editor, Ai Search Screa, AI video editor, AI video editor, AI research scira. In addition, he works with a robotics company to keep visual memories captured by a robot.
Although it is a slope towards consumers, the application is more like a playground for developers to understand the tool more and potentially use it in their workflows or their own applications.
Supermemory has important competitors in memory space. Startups like Lelicis supported by the Letta and MEM0 ventures (where Shah worked for a short time) build a layer of memory for agents. Susa Ventures de Supermemory, Susa Ventures, has invested in Memories.a with Samsung, who can use thousands of hours of sequences to obtain information. Shah says that if these startups could serve different industries and use cases, Supermemory will stand out because it offers lower latency.
“More and more AI companies will need a layer of memory. The Supermemory solution offers high performance while allowing you to quickly surface the relevant context. ” Said Browder.
