Taranjeet Singh (shown above, right) has founded six startups, experiencing both setbacks and varying levels of achievement. His latest venture, Mem0, may prove to be his most significant yet.
The company is built on the idea that large language models lack the ability to recall previous conversations like people do. If two individuals are interrupted during a chat, they can pick up where they left off. In contrast, AI models lose all context and must begin anew.
Mem0 addresses this gap. Singh describes it as a “memory passport,” allowing your AI memory to move with you between different apps and agents, much like your email or login credentials do now. The YC-supported company, which debuted in January 2024, has secured $24 million in funding ($3.9 million in previously undisclosed seed investment and a $20 million Series A).
The Series A was led by Basis Set Ventures, a fund focused on AI, with continued support from existing investors Kindred Ventures (which led the seed round) and Y Combinator, as well as new participants like Peak XV Partners and the GitHub Fund.
Prominent angel investors include Dharmesh Shah (HubSpot), Scott Belsky (former Adobe CPO), Olivier Pomel (Datadog), Thomas Dohmke (ex-GitHub CEO), Paul Copplestone (Supabase), James Hawkins (PostHog), Lukas Biewald (Weights & Biases), Brian Balfour (Reforge), Philip Rathle (Neo4j), and Jennifer Taylor (former Plaid president).
The fact that several influential figures in the software industry are backing Mem0 (pronounced “mem zero”) highlights its potential, and the momentum from its four-person team supports this view.
To date, the open source API, which claims to be the most widely used memory framework among AI developers, has earned over 41,000 stars on GitHub and has been downloaded more than 13 million times as a Python package. In the first quarter of 2025, Mem0 handled 35 million API requests. By the third quarter, that figure soared to 186 million, reflecting about 30% monthly growth.
In addition to its open source traction, over 80,000 developers have registered for its cloud platform. Mem0’s cloud API now processes more memory-related operations than any competitor and is the exclusive memory provider for AWS’s new Agent SDK.
At the start of 2023, Singh was still based in Bangalore, India. He began his professional journey as a software engineer at Paytm, one of India’s top startups, before becoming the first growth engineer at Khatabook. He left in late 2022, just as the ChatGPT surge was beginning, and went on to create one of the earliest GPT app marketplaces, which attracted over a million users.
This experience inspired him to launch Embedchain, an open source tool enabling developers to index, access, and synchronize unstructured data. As Embedchain gained traction, amassing over 8,000 GitHub stars, Singh sent more than 200 unsolicited emails to founders, investors, and engineers in Silicon Valley.
“I contacted nearly every well-known tech founder you can think of and was extremely persistent. Some replied, and after listening to our pitch, arranged for us to fly from Bangalore to San Francisco within 36 hours,” Singh recalled.
After arriving in the U.S., Singh reunited with his longtime friend, now co-founder and CTO, Deshraj Yadav, who previously led the AI Platform for Tesla Autopilot. Together, they had earlier developed EvalAI, an open source alternative to Kaggle that reached 1,600 GitHub stars.
While working on Embedchain, the pair also launched a meditation app inspired by Indian yogi Sadhguru. The app quickly gained popularity in India, but Singh notes that users repeatedly gave the same feedback: “I’m on this meditation journey, but the app doesn’t remember my progress.” This prompted them to shift their focus from Embedchain to Mem0 to address this issue.
The concept of memory in AI isn’t new, but it’s rapidly becoming a key area of competition. For example, OpenAI began piloting long-term memory features in ChatGPT in early 2024, and CEO Sam Altman has suggested that persistent memory will be a major feature of OpenAI’s forthcoming hardware. Other AI organizations are also experimenting with memory systems for their agents.
Singh contends that while major AI labs are developing memory solutions, they have little motivation to make them transferable or compatible with other systems. “Memory is turning into one of their main competitive advantages now that LLMs are becoming more standardized,” he explained.
He points out that while users can benefit from ongoing, tailored experiences in ChatGPT, developers aiming to build apps—such as a financial assistant that remembers a user’s trading history—require an open, neutral platform like Mem0.
“Our goal is for developers to deliver instant personalization through a shared memory network,” Singh said. “Picture it as Plaid, but for memory. That’s our next phase. For now, we’re dedicated to creating the best memory product possible.”
Mem0’s platform enables developers to save, access, and update user memory across different models, apps, and platforms. It works with any model—OpenAI, Anthropic, or open source LLMs—and integrates seamlessly with frameworks such as LangChain and LlamaIndex.
With Mem0, developers can build apps that become more intelligent with each use: therapy bots that recall previous sessions, productivity tools that remember user routines, and AI assistants that adapt over time. Their customers include solo developers as well as enterprise teams building copilots and automation solutions.
“We supported Mem0 from the very beginning—even before YC—because memory is essential for AI’s future,” said Lan Xuezhao, founder and partner at Basis Set Ventures. “We’re increasing our commitment as the team continues to address one of the toughest and most crucial infrastructure problems: enabling AI to develop lasting, contextual memory.”
Other emerging startups in the AI memory field include Supermemory (whose founder previously worked at Mem0), Letta (backed by Felicis), and Memories.ai.
