19-YO Indian American entrepreneur builds Supermemory, an open-source evaluation framework for all context systems, across any benchmark

Nineteen-year-old Indian American entrepreneur Dhravya Shah has developed Supermemory, a standard, open-source evaluation framework for all context systems, across any benchmark. Easily run and compare multiple memory providers, the Supermemory batteries included come with a Web UI, CLI, checkpoints and everything else.

Who is Dhravya Shah?

An IIT Bombay dropout, Shah is a Bachelor’s in Computer Science from Arizona State University. He started building Supermemory in Mumbai, while balancing preparations for the demanding IIT entrance exams with his early tries in consumer-facing apps and bots.

This young entrepreneur sold a Twitter formatting bot to Hypefury and then went to Arizona State University to pursue bigger goals.

Shah claims that Supermemory can assist a video editor to immediately find the perfect clip by prompt, or assist a real estate startup to sift through months of files for essential insight.

Supermemory has already drawn attention from developers and high-profile clients.

“It is fully open-source. You can view the results live in a web UI. Configure the answering/judge prompts. Restart from checkpoints. And view failed results, and exactly why they failed,” Dhravya explained on the social media platform X.

He added that they are doing a launch week in the holiday season for a specific reason.

“While other teams sleep, we are going to constantly ship every single day for the next 7 days,” the San Francisco-based young founder wrote on social media. “The last 7 days of 2025 is going to be UNFORGETTABLE,” he said.

Introducing memorybench by @supermemory 🔬

The standard, open source evaluation framework for all context systems, across any benchmark.

Easily run and compare multiple memory providers. batteries included - comes with a Web UI, CLI, checkpoints and everything else. pic.twitter.com/ORFEK23Shm

— Dhravya Shah (@DhravyaShah) December 24, 2025

All about Supermemory

The problem Dhravya Shah-led Supermemory is intending to solve is self-learning context about things (users, tasks, teams, etc.). Supermemory serves as a universal memory layer, allowing developers and users to add memory to their own large language models, for whatever the task might be.

The layer that solves this problem has to be independent of the model provider. This is because memory switching between providers would not only mean a lot of work for developers and users, but also would need support from these providers themselves.

If Google releases the next best model this week, but the user is stuck with OpenAI because their API has memory, they would be locked into using what they have used so far.

Memory should be a universal right, not a moat, as stated by Supermemory’s official page. Therefore, a layer that handles memory must exist, it reads. Supermemory acts as this independent layer. 

Memory needs to be semantic, just like the human brain that adapts, learns, and forgets. It remembers when it needs to. Something that is configurable and works for a wide variety of use cases, something that scales as all your users do, and thinks and infers things no one would have expected. Moreover, it needs to be fast, as you really don’t want personalisation to affect an experience, but to enhance it.

Intelligence without memory is nothing but sophisticated randomness. “We add memory to the intelligence,” claims Supermemory.

“One day, when AGI is a thing and robots are walking around everywhere, they will need a memory as sophisticated as their intelligence. And it would be super memory,” states the company page.

Also read: HN Reliance Foundation Hospital launches India’s first tele-robotic surgery programme (startuppedia.in)