Six active projects
LocoLLM, LocoBench, LocoConvoy, LocoEnsayo, LocoAgente, LocoPuente — each a distinct research question, all running on consumer hardware at Curtin University.
LocoLab
The work is deliberate. The hardware is secondhand. The questions are serious.
Local AI as a different bet, not a smaller one.
Mapping the floor
Most AI research documents the ceiling. We document the floor — what local small models can actually do on modest hardware — because most people live there and nobody is doing it honestly.
Local and private by design
Data sovereignty, individual privacy, institutional governance. Local inference is not a workaround for policy constraints — it is the only fully compliant path for many organisations, and the only honest answer for anyone who doesn’t want their data on someone else’s machine.
Five principles, one bridge
Engineer before hardware. Specialize and harness. Vary, don’t average. Conversation, not delegation. And LocoPuente — the bridge that brings frontier-equivalent tools onto local infrastructure.