Convergence, a London-based startup, has raised $12 million in pre-seed funding to further develop its personal AI agents. These agents can learn using long-term memory, much like humans do. The startup’s AI technology has the potential to revolutionize the way we work and live.
Convergence was founded by machine learning engineers Marvin Purtorab, CEO, and Andy Toulis, CTO. The duo met while working at Shopify and later joined enterprise AI startup Cohere in 2023. In April 2024, they founded Convergence, assembling a team of Google DeepMind, Meta, and OpenAI alumni.
Convergence’s Proxy AI agent has been developed for both employees and consumers. For employees, it can automate administrative and repetitive tasks and workflows to save time and increase productivity. For consumers, it can assist with everyday activities, such as booking trips or ordering groceries.
The Proxy AI agent’s differentiating factor lies in its ability to acquire skills by using long-term memory and continuous learning. According to Purtorab, “Our models have the ability to manage their own memories, decide to add things to their memories during inference time, and also delete things from their memory.”
Convergence’s AI technology is built on Large Meta Learning Models (LMLMs). LMLMs are a new generation of advanced machine learning models, built upon the principles of meta-learning. The startup is launching its Proxy agent in beta today, with a limited number of spots available.
The fresh capital will be used to develop new models to power its Proxy assistants. London-based VC firm Balderton Capital led the funding round with participation from Salesforce Ventures and Shopify Ventures.