Amid the race in artificial intelligence, large models have made significant progress in logical reasoning and multimodal interaction, yet they still face a尴尬 shortcoming—“forgetting immediately after turning away.” Recently, a new startup named Shenzhen Yiji Yuan Technology Co., Ltd. (MemoraX AI) has emerged, officially announcing a seed round funding of millions of dollars. This round was led by L2F Lightsource Venture Fund and Zhongding Capital, with several well-known investors also participating.

Notably, MemoraX AI is located in Nanshan, Shenzhen. From its establishment to securing this substantial investment, it took less than a month.

image.png

Industry Leader: A "Dual-Hybrid" Expert from Academia to the Frontline of Industry

He Jianye, the leader of MemoraX AI, is a typical "scholar entrepreneur." This 85-year-old professor has an extremely strong academic background and once conducted postdoctoral research at the Massachusetts Institute of Technology (MIT). After returning to China in 2015, he established one of the earliest deep reinforcement learning laboratories in China at Tianjin University.

Before founding MemoraX AI, He Jianye worked at Huawei for many years, serving successively as the director of Huawei's Decision Reasoning Laboratory, the director of the Large Model Algorithm Laboratory, and the Chief Technology Officer of the Medical Group. As the chief expert in Huawei's decision intelligence field, he led multiple billion-scale industrial projects. This resume, combining top-tier academic vision with industrial implementation experience, gave MemoraX AI a "big company" gene and practical characteristics right from the start.

Technical Breakthrough: Giving AI "Long-Term Memory"

Currently, intelligent agents often suffer from unstable information storage and inaccurate retrieval, leading to user experience gaps. The core logic of MemoraX AI lies in integrating memory capabilities deeply into the model's underlying structure through its self-developed "Agentic RL" technology, rather than simply attaching it to an external database.

According to the information, this technology mainly solves three dimensions of challenges:

  1. Dynamic Evolution: Memory is no longer a rigid pile of data but continuously understood, updated, and reconstructed during conversations.

  2. Precise Recall: In related memory tests, its performance significantly outperforms similar solutions, and training efficiency has increased hundreds of times.

  3. Scenario Reusability: Whether in daily office work, code writing, or digital companionship scenarios, memory can be seamlessly migrated.

Business Vision: Creating "Tailored" Digital Partners

From the perspective of the founding team, if AI cannot bridge the gap between "storage" and "memory," it will always remain just an efficient search engine. The goal of MemoraX AI is to evolve AI from a cold tool into a warm companion.

Currently, the company's business landscape has begun to take shape. On the B-side, MemoraX AI plans to provide standardized memory modules for professional fields such as finance, healthcare, and law, solving efficiency pain points caused by repeated inquiries; on the C-side, it aims to create personalized intelligent assistants that truly understand user preferences, learning habits, and work requirements. It is reported that the first batch of standardized memory products is expected to be officially launched within the next year.