01
Tools
We build frameworks and utilities for training and experimenting without a hyperscale stack, including Ellipse in pure C.
About
We study what it would take for machines to reach human-level intelligence, and we build products people can actually use, including where computers and connections are modest.
01
We build frameworks and utilities for training and experimenting without a hyperscale stack, including Ellipse in pure C.
02
We build and release benchmarking data so Swahili and other African languages can be evaluated fairly, sector by sector.
03
We train models and put them where people can try them, build on them, and run them themselves.
04
We build the engines that let those models run on ordinary machines, including when the computer is a CPU and the network is thin.
05
We publish evaluations so anyone can see what a model actually understands, and compare the next one fairly.
Research becomes products people can use: local, voice, language, and visual intelligence under one brand.
01
Local intelligence
Offline-first assistant that keeps models and data on your own machine, built for thin networks and private use.
02
Voice intelligence
Voice platform for Kiswahili first: talk, speak, and listen with conversational AI and speech tools.
03
Language intelligence
Text platform for African languages: chat, translation, and a citation-backed African knowledge layer.
04
Visual intelligence
Visual understanding of images and video, and tools to generate and edit visuals.
Roles and internships.
Maliga Road, Goba, Dar es Salaam, Tanzania