Farstead Systems
Farstead builds data and AI infrastructure for the languages no model speaks and the places no network reaches, starting in Africa, starting with Tiv.
Thesis
The largest opportunities in African technology aren't in copying what already works elsewhere. They're in the gaps everyone steps over, the languages no model speaks, the regions no signal reaches, the infrastructure that's assumed but absent.
We build there. It's slower, harder, and unglamorous. It is also where the durable businesses are made, anchored to specific, quantifiable problems for real people in real conditions, not to whatever is fashionable this year.
What we build
Seven thousand languages in the world. Fewer than a hundred that modern AI can meaningfully speak. We build the corpora, models, and applied tools that close that gap, beginning with Tiv, spoken by over five million people and all but absent from the digital world.
Where power flickers and networks drop, ordinary infrastructure quits. We design communication and edge-intelligence systems that keep working in the conditions most technology is never tested against.
Products — nodes in the network
The first open, community-sourced AI for the Tiv language.
Building the corpus, models, and applied tools to bring AI to five million Tiv speakers, open where others stay closed, voice-bound where others stop at text, deployed across health, agriculture, and government.
kwaghfan.comSecure, adaptive mesh communication.
A mesh layer that runs on phones, our own nodes and the radios you already have. It carries messages and reaches local services without the internet, seals every message, and adapts on its own when links fail or are jammed.
See IndyrServices
We offer consulting, field deployments and full training programmes in the areas we work in every day. We only offer what we have built and tested ourselves.
We design and install LoRa mesh networks on large farms, linking sensors on water tanks, soil, pumps and irrigation. You can see conditions across the whole farm and switch equipment on and off remotely, without mobile coverage.
We field-test LoRa mesh networks on your site and report on range, link quality and delivery under real conditions. We also design LDPC frames and test their error-correction performance in Python and MATLAB before they reach your firmware.
We build AI assistants that work on your own documents and data, run on infrastructure you control, and are designed to keep sensitive information inside your business. We also build language tools for African languages.
Training programmes
We train and grow local engineers, because the people closest to the problem understand it best.
A full programme in building, deploying and testing low-power wireless systems, covering microcontroller firmware, LoRa and LoRaWAN, mesh protocols, low-power design and field testing. You finish by deploying a working network on a real farm.
A full programme in building AI systems for businesses, covering retrieval over company documents, self-hosted open models, evaluation and data protection. You finish with a working assistant built on real data.
Each programme runs part-time over 12 weeks in small cohorts, with a capstone project you can show employers. The first cohort starts in mid-January 2027.
Partner, funder, researcher, or someone who wants to help bring technology to where it hasn't reached, we'd like to hear from you.