At just nineteen, Indian-American entrepreneur Dhravya Shah has created  Supermemory, an innovative open-source evaluation framework aimed at standardising  the benchmarking and comparison of context and memory systems across various  platforms. This framework enables developers to efficiently test multiple memory  providers with a single setup, featuring built-in tools such as a web user interface,  command-line interface (CLI), checkpoints, and comprehensive evaluation reports. 

Supermemory seeks to fill a significant void in contemporary AI systems by ensuring  consistent, portable, and intelligent memory across different models and providers. 

Who Is Dhravya Shah? 

Shah, who left IIT Bombay to pursue a Bachelor’s degree in Computer Science at  Arizona State University, embarked on his entrepreneurial journey early on. He began  developing Supermemory while living in Mumbai, balancing his preparation for the  highly competitive IIT entrance exams with experimentation on consumer-focused  applications and bots. 

Prior to his move to the US, Shah successfully built and sold a Twitter formatting bot to  Hypefury, which not only bolstered his confidence but also provided the financial  backing to explore more ambitious projects. Now based in San Francisco, he is focused  on scaling Supermemory. 

What Supermemory Does 

Shah explains that Supermemory can facilitate a variety of real-world applications. For  example, a video editor can swiftly retrieve the most pertinent clip with a simple 

prompt, while a real estate startup can efficiently analyse months’ worth of stored  documents to extract vital insights. 

The platform has begun to attract attention from developers and several prominent  clients within the AI ecosystem. 

Highlighting the product’s transparency and flexibility, Shah shared on social media  platform X, “Supermemory is fully open-source. You can view results live in a web UI,  configure judge prompts, restart runs from checkpoints, and see exactly why certain  results failed.” 

Launch Strategy and Vision 

Shah has strategically timed the launch of Supermemory for the holiday season. “While  other teams rest, we are committed to shipping every single day for the next seven  days,” he stated. Describing this as an intense release cycle, he promised that the final  week of 2025 would be “unforgettable.” 

Addressing the Memory Challenge in AI 

At its essence, Supermemory is designed to tackle the issue of self-learning context— memory associated with users, tasks, workflows, and teams. It serves as a universal  memory layer that can be integrated into any large language model, regardless of the  underlying provider. 

Shah contends that memory should not be confined to a specific model or company.  The current complexity and restrictions of switching memory systems between  providers often trap users within a single ecosystem. “Memory should be a universal  right, not a moat,” the Supermemory website asserts. 

By being provider-agnostic, Supermemory allows developers to adopt superior models  without losing their accumulated context. 

Building Memory Like the Human Brain 

Supermemory is crafted to function semantically, akin to the human brain— remembering what is essential, forgetting when necessary, adapting over time, and  scaling with usage. It is designed to be configurable, fast, and intelligent, ensuring that  personalisation enhances the user experience rather than hindering it. 

“Intelligence without memory is just sophisticated randomness,” the company states.  “We add memory to intelligence.” 

Looking to the future, Shah envisions that advanced AI systems—and even robots—will  require memory systems as robust as their intelligence. “When AGI becomes a reality, it  will need a super memory,” the company notes.