MemoraX AI
Category: Foundation Models / LLMs
MemoraX AI raises $10M seed to embed long-term memory into LLM weights via Agentic RL, targeting enterprise knowledge management and personalized AI agents. MemoraX AI was founded in 2026. The company is led by Hao Jianye. Based in Shenzhen, China. Total funding raised: $13.8M+ (100M+ RMB). Latest round: Seed. Key investors include L2F Light Source Entrepreneurs Fund, Zhongding Capital, CICC Capital, Datai Capital, Huaye Tiancheng.
- Founded
- 2026
- Headquarters
- Shenzhen, China
- Total funding
- $13.8M+ (100M+ RMB)
Value proposition
Endogenous memory architecture directly embedded in model weights, eliminating reliance on external retrieval systems, enabling persistent, self-updating memory for agentic AI.
Products and solutions
Endogenous Memory Module
Target customer
Enterprises needing knowledge management, personalized AI agents, and consumer AI companions
Industries served
Enterprise Knowledge Management, Consumer AI Assistants, AI Companions
Technology advantage
Agentic RL (reinforcement learning) technology that trains memory directly into LLM weights; proven industrial track record in RL deployments (chip design, autonomous driving, gaming); team combines top academic RL talent with large-scale industry deployment experience from Huawei, Alibaba, Tencent.
How they differentiate
MemoraX embeds memory directly into model weights through Agentic RL, eliminating the need for external retrieval systems (e.g., RAG, vector databases) and enabling persistent, self-updating memory that can transfer across scenarios.
Key partnerships
Oxford University (co-developed ScriptMem and SWE-context-bench benchmarks), Nanjing University & Shanghai AI Lab (LoCoMo-Refined benchmark), TusStar (启迪之星) incubation
Major milestones
Founded March 2026, Raised $10M seed round co-led by L2F Light Source Entrepreneurs Fund and Zhongding Capital (April 2026), Achieved state-of-the-art memory performance on LoCoMo-Refined benchmark (30% ahead of 2nd place), Raised Seed+ round (tens of millions RMB) led by CICC Capital, Datai Capital, Huaye Tiancheng (May 2026), Cumulative funding exceeds 100M RMB, 10 papers accepted at ICML 2026, Co-developed ScriptMem and SWE-context-bench benchmarks with Oxford University
Market positioning
MemoraX AI raises $10M seed to embed long-term memory into LLM weights via Agentic RL, targeting enterprise knowledge management and personalized AI agents.
About Hao Jianye
Professor at Tianjin University (Elite Professor, National Outstanding Youth Scientist); Former Director of Decision & Inference Lab at Huawei, Director of Large Model Algorithm Lab at Huawei, Technology VP of Huawei Medical Legion; Chief expert in decision intelligence at Huawei; Google Scholar citations 15,000+; Top 10 global research output at ICML/NeurIPS/ICLR
Official website: https://memorax.ai