19-Year-Old’s AI Memory Startup Supermemory Wins Backing From Google Execs
▼ Summary
– AI models have limited long-term memory across sessions, prompting researchers to develop new memory solutions.
– Dhravya Shah created Supermemory, a universal memory API that builds knowledge graphs from unstructured data to personalize context for AI applications.
– Supermemory can ingest various data types including files, chats, and emails, and offers features like a chatbot, notetaker, and Chrome extension.
– The startup secured $2.6 million in seed funding from investors including Susa Ventures and individuals from Cloudflare, Google, and OpenAI.
– Supermemory faces competition from other memory layer startups but claims differentiation through lower latency and broad applicability across industries.
The challenge of equipping artificial intelligence with long-term memory is a central focus for developers aiming to create more intuitive and useful applications. While AI models have seen their context windows, the amount of information they can process in a single session, grow significantly, they typically cannot retain context across multiple interactions. A new startup, Supermemory, founded by 19-year-old Dhravya Shah, is tackling this very problem by developing a sophisticated memory solution designed for AI applications.
Shah, originally from Mumbai, began his entrepreneurial journey by creating consumer-focused bots and applications. He successfully sold one of his early creations, a bot that transformed tweets into visually appealing screenshots, to the social media platform Hypefury. The proceeds from this sale provided him with the financial freedom to alter his educational plans. Instead of preparing for the Indian Institute of Technology entrance exams, he moved to the United States to attend Arizona State University.
Once settled in the U.S., Shah embarked on a personal challenge to build a new project every week for forty weeks. It was during this intensive period that the initial version of Supermemory, then named Any Context, was born and shared on GitHub. This early iteration was a tool that enabled users to converse with their saved Twitter bookmarks.
The platform now functions as a universal memory API, constructing a detailed knowledge graph from processed data to deliver a personalized context for each user. It can, for example, help a writing application query entries from several months prior or assist an email client in searching through old messages. Because it accepts multimodal inputs, the technology could also allow a video editor to pull relevant clips from a vast library based on a simple text prompt.
According to the company, Supermemory can ingest a vast array of data types, including documents, chat histories, project files, emails, PDFs, and live application data streams. Its integrated chatbot and notetaker feature lets individuals add memories in text form, upload files or links, and connect to popular services like Google Drive, OneDrive, and Notion. A dedicated Chrome extension further simplifies the process of capturing notes directly from websites.
“Our fundamental advantage lies in our ability to pull meaningful insights from any form of unstructured data, thereby giving applications a richer understanding of the user,” Shah explained. “Since we operate across multimodal data streams, our solution is adaptable to all kinds of AI apps, from email clients to advanced video editors.”
Supermemory has successfully raised $2.6 million in a seed funding round. The investment was led by Susa Ventures, Browder Capital, and SF1.vc, with participation from prominent individual backers. These include Cloudflare’s CTO Dane Knecht, Google’s AI lead Jeff Dean, DeepMind product manager Logan Kilpatrick, Sentry founder David Cramer, and executives from OpenAI, Meta, and Google. Shah noted that the prestigious accelerator Y-Combinator also expressed interest, but the timing did not align as he had already secured investors.
Joshua Browder, the founder and CEO of the “robot lawyer” startup DoNotPay, who operates Browder Capital, was particularly impressed by Shah’s drive. “I connected with Dhravya on social media, and his incredible speed in building and shipping products was what convinced me to invest in him,” Browder stated.
The startup already boasts a growing list of customers, such as the a16z-backed desktop assistant Cluely, AI video editor Montra, search tool Scira, Composio’s Rube, and real estate platform Rets. It is also collaborating with a robotics company to help robots retain and utilize visual memories. While the application has consumer-friendly elements, it primarily serves as a testing ground for developers to explore its capabilities for integration into their own workflows and products.
The AI memory space is becoming increasingly competitive, with several other startups like Letta and Mem0, where Shah briefly worked, also developing memory layers for AI agents. Even Supermemory’s own investor, Susa Ventures, has backed a rival, Memories.ai. Shah believes his company will distinguish itself by offering superior performance, specifically through lower latency, which ensures relevant context is surfaced almost instantly.
Browder echoed this sentiment, adding, “The demand for a reliable memory layer in AI is only going to increase. Supermemory’s solution delivers high performance while allowing you to surface relevant context quickly, which is a critical advantage.”
(Source: TechCrunch)