NeoCognition, a research lab focused on building self-learning AI agents, has emerged from stealth with a $40 million seed funding round to advance specialized intelligence and expert-level automation. The oversubscribed round was co-led by Cambium Capital and Walden Catalyst Ventures, with participation from Vista Equity Partners and several other investors. Notable angel investors and advisors include Lip-Bu Tan, Ion Stoica, and leading AI researchers such as Dawn Song, Ruslan Salakhutdinov, and Luke Zettlemoyer.
Founded by Yu Su, along with co-founders Xiang Deng and Yu Gu, NeoCognition is rooted in years of academic research conducted at The Ohio State University. Su’s team has been working on large language model-based agents well before the rise of generative AI platforms, contributing foundational research such as Mind2Web, MMMU, and SeeAct. Their work has influenced leading AI systems developed by OpenAI, Anthropic, and Google.
NeoCognition is focused on developing a new class of AI agents that continuously learn from their environments, enabling them to specialize into domain experts over time. By building structured “world models” of workflows and constraints, these agents aim to become faster, more cost-effective, and more reliable, particularly in high-stakes applications where precision and safety are critical.
The company positions its approach as a departure from today’s general-purpose AI agents, emphasizing continuous learning and specialization as the path toward expert-level intelligence. Its platform is designed to reduce the need for manual customization while improving performance in complex, real-world tasks.
Headquartered in Palo Alto, California, NeoCognition enters the market with $40 million in committed seed capital and backing from a network of prominent investors and researchers, signaling strong early confidence in its approach to building the next generation of AI agents.
KEY QUOTES:
“Dr. Su and his team have already developed research that spans every piece of the agent puzzle, ranging from perception to memory, planning, evaluation, and safety. We are confident NeoCognition is uniquely positioned to tackle the hardest challenges in agentic AI.”
Lip-Bu Tan, CEO, Intel; Founding Managing Partner, Walden Catalyst Ventures
“AI today is fundamentally unreliable when it comes to executing real work that requires deep expertise. The true power of human intelligence is the ability to continuously learn and specialize. Our approach mirrors how humans gain expertise on the job through building a structured model of their micro-world, and would eliminate the extensive manual customization required by current models.”
Yu Su, CEO and Co-Founder, NeoCognition
“As general-purpose agents become table stakes, the key challenge in AI is achieving expert-level intelligence. NeoCognition’s new approach to building agents that learn to become experts has the potential to reach the level of reliability, efficiency, and cost-effectiveness required for high-stakes applications.”
Ion Stoica, Co-Founder, Databricks; Professor, UC Berkeley
“At the core of NeoCognition is a novel learning mechanism that will allow agents to specialize very quickly. We have strong conviction in the team’s expertise and believe their research is charting a new path toward specialized intelligence that will democratize access to frontier agent capabilities.”
Landon Downs, Managing Partner, Cambium Capital
