Summary:
19-year-old Dhravya Shah founded Supermemory, an AI memory startup with backing from Google executives.
Supermemory addresses long-term memory issues in AI models by extracting insights from unstructured data.
The startup evolved from a GitHub project to a full-time venture after Shah's experience at Cloudflare.
It offers a universal memory API that builds knowledge graphs and supports multimodal inputs for apps.
Key features include a chatbot, notetaker, and integrations with tools like Google Drive and Notion.
The Rise of Supermemory: A Teen Founder's Vision
Context windows in AI models, which determine how much information a model can "remember," have expanded over time. Yet, researchers have suggested new approaches to enhance long-term memory, as AI often struggles to retain context across multiple sessions.
Nineteen-year-old Dhravya Shah is tackling this challenge head-on with his startup, Supermemory, a memory solution designed for AI applications. Hailing from Mumbai, India, Shah began his journey by creating consumer bots and apps, even selling a tweet-formatting bot to Hypefury. This success funded his move to the U.S., where he enrolled at Arizona State University.
Image Credits: Supermemory
Shah embarked on a 40-week challenge to build something new each week, leading to the creation of Supermemory (initially called Any Context) on GitHub. It started as a tool for chatting with Twitter bookmarks but has evolved into a sophisticated system that extracts memories and insights from unstructured data, improving contextual understanding for apps.
Image Credits: Supermemory
After an internship and role at Cloudflare, where he focused on AI and infrastructure, Shah was encouraged by advisors, including Cloudflare CTO Dane Knecht, to productize Supermemory. This year, he committed to building it full-time.
How Supermemory Works
Supermemory is now a universal memory API for AI apps, constructing a knowledge graph from processed data to personalize user context. It supports queries across old entries for apps like writing tools or email clients and handles multimodal inputs, enabling tasks such as fetching relevant video assets from a library.
The startup ingests diverse data types, including files, documents, chats, projects, emails, PDFs, and app data streams. Features include a chatbot and notetaker for adding memories via text, files, or links, with integrations for Google Drive, OneDrive, and Notion. A Chrome extension simplifies note-taking from websites.
"Our core strength is extracting insights from any unstructured data and giving apps more user context. As we handle multimodal data, our solution fits various AI apps, from email clients to video editors," Shah explained.
Comments