MemoraX AI Memory Model Raises Million-Dollar Seed Round to End AI Amnesia
Can large models truly serve as efficient "search engines"? Hao Jianye, founder of Shenzhen Yijiayuan Technology Co., Ltd. (MemoraX AI), argues that without bridging the gap between "storage" and "memory," AI will never become a genuine intelligent partner.
Founded just a month ago, this startup has announced a $10 million seed round led by L2F Lightsource Entrepreneur Fund and Zhongding Capital, with participation from notable investors and industry players. The funding will primarily support the iteration of Agentic RL (agent reinforcement learning) algorithms, engineering implementation, and the development of internal memory modules.

MemoraX AI was established in March 2026. Its founder, Hao Jianye, is a distinguished professor at Tianjin University who previously served as Director of Huawei's Decision Reasoning Lab and Large Model Algorithm Lab. He brings strong academic credentials and hands-on experience in reinforcement learning. The team consists of technical experts from major companies like Huawei, Alibaba, and Tencent, forming a robust blend of academic and industry expertise.
On the technical side, MemoraX AI uses its proprietary Agentic RL technology to internalize memory. This approach addresses common pain points of current large models, such as fragmented memory and imprecise retrieval. Its core advantages span three dimensions: dynamic evolution—memory continuously updates and reorganizes during interactions rather than remaining static; precise recall—outperforming peers by 30% on the LoCoMo-Refined test set while achieving 400x training efficiency; and generalization and reuse—adapting flexibly across scenarios from smart terminals to enterprise-level knowledge management.
These cutting-edge technologies are already being applied in fields such as autonomous driving, chip design automation, and industrial solvers.
Commercially, MemoraX AI pursues a dual strategy targeting both B2B and B2C markets. On the B2B side, it offers standardized memory modules for industries like healthcare, finance, and law, enhancing intelligent customer service and knowledge management. On the B2C side, it focuses on creating more personalized digital companions. The first standardized memory products are expected to officially launch within a year.
Investors believe that a memory system is the most fundamental infrastructure for an agent, directly determining the upper limit of AI delivery capabilities. As the AI application layer urgently needs to overcome bottlenecks, MemoraX AI's approach of self-updating through dynamic interactions is seen as a key effort to push past the industry's ceiling.
Related article
Xiaoyunque AI Unveils Short Drama Agent Powered by Seedance 2.0 for One-Click Film Conversion
On March 19, ByteDance’s all-in-one AI content creation platform Little Swallow AI officially unveiled its “Short Drama Agent” feature. Recognized as the industry’s first intelligent agent powered by the Seedance 2.0 algorithm, this tool automates th
NanoClaw Creator Rejects $20M Buyout, Secures $12M Seed Funding
NanoCo, the team behind NanoClaw—a security-first alternative to OpenClaw—has secured an oversubscribed $12 million seed round following a viral launch, founders Gavriel and Lazer Cohen told TechCrunch.Valley Capital Partners led the funding round, w
SpaceX Financial Report Shows AI Revenue Tripling, Surpassing Space Launch
SpaceX’s latest quarterly financial report reveals that its AI division, bolstered by substantial investments in computing infrastructure, has generated $2.6 billion in revenue for a single quarter. This figure represents more than a threefold year-o
Related Special Topic Recommendations
Comments (1)
0/500
Can large models truly serve as efficient "search engines"? Hao Jianye, founder of Shenzhen Yijiayuan Technology Co., Ltd. (MemoraX AI), argues that without bridging the gap between "storage" and "memory," AI will never become a genuine intelligent partner.
Founded just a month ago, this startup has announced a $10 million seed round led by L2F Lightsource Entrepreneur Fund and Zhongding Capital, with participation from notable investors and industry players. The funding will primarily support the iteration of Agentic RL (agent reinforcement learning) algorithms, engineering implementation, and the development of internal memory modules.

MemoraX AI was established in March 2026. Its founder, Hao Jianye, is a distinguished professor at Tianjin University who previously served as Director of Huawei's Decision Reasoning Lab and Large Model Algorithm Lab. He brings strong academic credentials and hands-on experience in reinforcement learning. The team consists of technical experts from major companies like Huawei, Alibaba, and Tencent, forming a robust blend of academic and industry expertise.
On the technical side, MemoraX AI uses its proprietary Agentic RL technology to internalize memory. This approach addresses common pain points of current large models, such as fragmented memory and imprecise retrieval. Its core advantages span three dimensions: dynamic evolution—memory continuously updates and reorganizes during interactions rather than remaining static; precise recall—outperforming peers by 30% on the LoCoMo-Refined test set while achieving 400x training efficiency; and generalization and reuse—adapting flexibly across scenarios from smart terminals to enterprise-level knowledge management.
These cutting-edge technologies are already being applied in fields such as autonomous driving, chip design automation, and industrial solvers.
Commercially, MemoraX AI pursues a dual strategy targeting both B2B and B2C markets. On the B2B side, it offers standardized memory modules for industries like healthcare, finance, and law, enhancing intelligent customer service and knowledge management. On the B2C side, it focuses on creating more personalized digital companions. The first standardized memory products are expected to officially launch within a year.
Investors believe that a memory system is the most fundamental infrastructure for an agent, directly determining the upper limit of AI delivery capabilities. As the AI application layer urgently needs to overcome bottlenecks, MemoraX AI's approach of self-updating through dynamic interactions is seen as a key effort to push past the industry's ceiling.
Xiaoyunque AI Unveils Short Drama Agent Powered by Seedance 2.0 for One-Click Film Conversion
On March 19, ByteDance’s all-in-one AI content creation platform Little Swallow AI officially unveiled its “Short Drama Agent” feature. Recognized as the industry’s first intelligent agent powered by the Seedance 2.0 algorithm, this tool automates th
NanoClaw Creator Rejects $20M Buyout, Secures $12M Seed Funding
NanoCo, the team behind NanoClaw—a security-first alternative to OpenClaw—has secured an oversubscribed $12 million seed round following a viral launch, founders Gavriel and Lazer Cohen told TechCrunch.Valley Capital Partners led the funding round, w
SpaceX Financial Report Shows AI Revenue Tripling, Surpassing Space Launch
SpaceX’s latest quarterly financial report reveals that its AI division, bolstered by substantial investments in computing infrastructure, has generated $2.6 billion in revenue for a single quarter. This figure represents more than a threefold year-o





Home






