NeoCognition Raises $40 Million Seed Round to Build Task-Learning Specialized AI Agent

According to Dynam.AI monitoring, the AI research lab NeoCognition has announced the completion of a $40 million seed round, emerging from stealth mode. Headquartered in Palo Alto, California, the company was co-founded by Ohio State University Associate Professor Yu Su, along with Xiang Deng and Yu Gu. This oversubscribed round saw Cambium Capital and Walden Catalyst Ventures co-leading, with participation from Vista Equity Partners. Angel investors include Intel CEO Lip-Bu Tan, Databricks co-founder and Chairman Ion Stoica, and AI researchers Dawn Song, Ruslan Salakhutdinov, and Luke Zettlemoyer.

NeoCognition aims to address the issue of reliability in intelligent agents. Co-founder Yu Su stated that currently, AI agents only succeed in their tasks as expected about 50% of the time, making each execution akin to rolling a dice. The company's approach involves enabling agents to quickly specialize in their roles by leveraging a self-built domain-specific "world model" that captures the rules, relationships, and constraints within a particular business environment, rather than relying on general pre-training. Prior to ChatGPT, Yu Su had established the LLM Intelligent Agents Research Lab at Ohio State University, and the team's previous representative work includes Mind2Web, MMMU, and SeeAct.

The company's business roadmap is directed towards the enterprise sector, primarily serving SaaS companies to embed continuously improving AI employees in their products and operational processes. Vista Equity Partners, which holds a portfolio of enterprise software companies, is seen as a distribution channel through this investment. With a team of around 15 members, most of whom hold doctoral degrees, the company currently does not have any publicly available products.