Locai Labs is a British research lab for local AI, based in London.
We work across the full local AI stack: training and optimising AI models, building the software needed to run them, and making those models practical on local hardware.
Our focus is on two core research problems.
The first is efficient inference: making advanced open-weight models smaller and faster through techniques including structured pruning, post-training quantisation and GPU-specific optimisation.
The second is continual learning: helping models keep learning from private data after deployment without losing what they already know.
Our main product is Locai One, an on-prem AI appliance designed to run AI for an organisation from inside its own network.