Unimatrix Labs supports companies building networks of AI agents that learn throughout their operational life and share knowledge across the collective, so the whole adapts faster than any single system could alone.
Today's large models learn mostly during one training phase and rarely change afterward. Knowledge stays locked inside the organisations that hold it. Unimatrix Labs takes a different route, grounded in research on lifelong and collective machine learning.
Independent units acquire new skills as they operate, then share what they learn across a network. A unit joins by connecting to one existing member, which passes on the addresses of others. Collectives form, grow, and reconfigure as units join or leave. Through federated and transfer learning, capability accumulates across the whole rather than being retrained from scratch in each system.
Each agent keeps its own objectives and independence. The aim is a network of cooperating systems that resists domination by a few large models, recycles knowledge to reduce energy demand, and adapts to new tasks in the field.
Grounded in research published in Nature Machine Intelligence: A collective AI via lifelong learning and sharing at the edge.
Both divisions specialise on decentralised, distributed collectives. They differ in how authority is arranged, which makes each suited to a different class of task.
Fully decentralised and distributed. Units coordinate peer to peer with no command centre, organising themselves through local interaction and knowledge sharing. A democracy of agents, each retaining its own goals.
Decentralised and distributed, but with a central authority that directs the collective. Knowledge still flows across the network, while coordination runs through a command centre for tasks that need oversight and accountable control.
Agents keep learning as they operate, adapting to new tasks and conditions instead of freezing after a single training run.
Units broadcast and absorb skills across the network, so a capability learned once can spread to the whole collective.
Every agent keeps its own objectives. Cooperation without homogenisation, modelled on healthy collective systems.
Unimatrix Labs brings together researchers and engineers working on lifelong and collective machine learning. Our team translates the science of decentralised, knowledge-sharing AI into systems that operate in the field.
Associate Professor in Computer Science. Leads the research direction on lifelong and collective learning.
Short biography placeholder. Replace with this person's background and contribution to the company.
Short biography placeholder. Replace with this person's background and contribution to the company.
For research collaboration, partnerships, and enquiries.