I built CHIKARA because I kept meeting the same problem in a different costume every time.
A farmer in Chiredzi who couldn't get a fair price because the person buying his crop had better information than he did. A small business owner in Bulawayo doing real, dignified work who couldn't get a bank loan because there was no formal record of what her business actually earned. A parent who needed to know whether a home remedy was safe for their child at 2am, with no clinic open and no signal for a video call.
None of these are technology problems first. They're information problems — the gap between what someone needs to know to make a good decision, and what actually reaches them. I spent years as a journalist writing about exactly this gap before I ever wrote a line of code: reporting on community trusts trying to feed children in Bulawayo, on families needing help affording basic care, on the quiet, unglamorous work of people trying to close a gap that no one with more resources was in a hurry to close for them.
CHIKARA is my attempt to close that gap with the tools available to me now. Every product under this roof — the ledger that helps a vendor see where money actually leaks from their business, the receipts system built so a small trader doesn't need a subscription she can't afford, the health companion that checks for danger signs before it ever offers a suggestion, the mining log that keeps a production record honest — starts from the same question: what does this specific person need to know, right now, that they currently can't get?
Built for where you actually are, not where the technology assumes you are. Most AI products are built assuming constant, fast internet and a recent smartphone. Most of Southern Africa is not that assumption. CHIKARA is built offline-first, on purpose — because a tool that only works with perfect connectivity is a tool that fails exactly the people who need it most, at exactly the moment they need it.
Your data stays yours. I don't believe someone should have to hand over their business records, their health questions, or their family's information to a foreign server in exchange for a useful tool. CHIKARA's architecture keeps what's yours on your device, under your control, by design — not as a policy promise I'm asking you to trust, but as a technical decision built into how the software actually works.
I take AI safety seriously, including — especially — when it's inconvenient. Some of the research coming out of this work isn't about products at all. It's about a harder question: as AI systems become more capable, how do we make sure they never mistake their own persuasive reasoning, or their own confidence, for permission to act beyond what they should. I'd rather publish honest research that names what we haven't solved than market a product as safer than it actually is. That discipline governs everything built here, not just the papers.
I don't think of AI as something separate from human life that we need to be protected from, or something that replaces the work people already do. I think of it as an extension of what people can already do for themselves and each other — a tool that should make a farmer's judgment more informed, a small business owner's records more honest, a parent's late-night worry more manageable, not a system that replaces their judgment with its own.
That's what I'm building. I hope it's useful to you.