By Aditya Agarwal 
A different route into entrepreneurship
At 19, Dhravya Shah is building an AI infrastructure company around a problem that has become increasingly important as AI systems become more capable.
The Mumbai-born entrepreneur is the founder of Supermemory.
The company focuses on memory infrastructure for large language models and AI agents.
Shah left college to pursue the company full-time.
His path differs from the conventional Indian technology route, where engineering graduates often prioritise placements before considering entrepreneurship.
However, his decision was shaped by earlier experimentation with AI tools and software development.
From small AI tools to infrastructure
Shah’s entrepreneurial journey began with smaller AI projects.
His experience at Cloudflare also exposed him to large-scale technology infrastructure.
That background influenced his thinking about AI.
Large language models are increasingly capable of reasoning.
However, they do not automatically retain useful context across applications and sessions.
That creates a problem.
An AI system may be intelligent but still lack durable knowledge about a user, company or workflow.
Supermemory is attempting to address that gap.
What AI memory actually means
AI memory is different from simply storing conversation history.
A useful memory system needs to identify relevant information.
It needs to retrieve that information when required.
It also needs to deal with changing facts.
Furthermore, it needs to work across different models and applications.
Supermemory describes its product as an interoperable memory layer for LLMs and agents.
The company has built infrastructure including its own vector database, content parser and extraction systems.
That makes the business an infrastructure play rather than another consumer chatbot.
The company has already raised $3 million
Supermemory announced a $3 million funding round led by Susa Ventures, Browder Capital and SF1.vc.
The round also included individual investors from companies and technology communities including Cloudflare and other AI infrastructure firms.
The company said its customers include enterprises and open-source projects.
Those customer relationships matter because infrastructure companies generally need developers to integrate their technology into production systems.
A developer who builds an application around a memory layer can potentially become a long-term customer.
That creates a different business model from consumer AI applications.
Why memory could become a major AI layer
The AI industry is moving toward agents.
Agents do more than answer questions.
They use tools.
They interact with applications.
They maintain workflows.
They also need context.
Without reliable memory, an agent can become repetitive or lose important information between interactions.
Memory therefore becomes a potential infrastructure layer.
It can store user preferences.
It can retain project knowledge.
It can connect information from multiple sources.
It can also provide relevant context to different AI models.
That makes the technology increasingly relevant as agent-based computing expands.
Supermemory changes direction
The company has also recently changed its product focus.
In September, Supermemory discontinued its company brain product and its Nova personal knowledge-management platform.
The company said users were refunded and that it was concentrating on its memory API, MCP and plugins.
That decision is revealing.
Instead of building several end-user applications, Supermemory is focusing on infrastructure.
For an early-stage company, such focus can help concentrate engineering resources.
It also clarifies the target customer.
The company is effectively positioning itself as a technology layer that other AI products can use.
The founder’s age is less important than the model
Shah’s age has naturally attracted attention.
However, the more significant entrepreneurial story is the company’s positioning.
AI infrastructure is becoming increasingly specialised.
There are companies focused on models.
Others build vector databases.
Some develop agent frameworks.
Supermemory is attempting to occupy the memory layer.
That creates a narrower but potentially important market.
The challenge will be differentiation.
AI companies can build memory systems internally.
Infrastructure providers therefore need to offer reliability, speed, interoperability and developer convenience.
The next phase
Supermemory’s shift toward infrastructure gives Shah a more focused business problem.
The company must convince developers that memory should be a separate infrastructure layer.
It must also demonstrate that its technology can work across models and applications.
If that happens, memory could become a standard component of AI-agent architectures.
For Shah, the journey therefore goes beyond being a young founder.
It is a test of whether a small infrastructure company can become part of the foundation supporting a much larger AI ecosystem.
Tags: Supermemory, Dhravya Shah, entrepreneurship, AI memory, LLMs, AI infrastructure, Indian founders
Flairius CTA: Follow Flairius News — sharp takes on AI, business, and India’s startup economy — flairiusnews.com