01 / RESEARCH
Our work sits where model efficiency, local relevance and open science meet.
01
Researching compact language models that deliver useful intelligence with radically lower compute, memory and deployment costs.
We are interested in the full efficiency stack: model architecture, data quality, training strategy, inference, compression and deployment on constrained hardware.
02
Building methods and datasets that help AI understand the languages, knowledge and contexts that global systems routinely overlook.
Our goal is not to place a regional wrapper around general models. It is to investigate how local knowledge and language can shape data, evaluation and system design from the beginning.
03
Making models easier to inspect, adapt and trust through open research, transparent evaluation and practical safety work.
We want openness to be useful, not performative. That means clear methods, reproducible evaluation and honest reporting of both capability and limitation.
FLAGSHIP PROGRAMME
In development
Building a family of compact, capable language models designed for private, affordable and locally adaptable intelligence.
BUILD WITH US
Researchers, engineers, institutions and thoughtful funders: if this mission speaks to you, we would like to hear from you.
Start a conversation