The trusted third-party
unlocking agricultural finance
in West Africa.

Across West Africa, more than 60 million head of cattle walk outside the formal financial system. Herders hold real wealth and yet cannot borrow against it, insure it, or sell it at a fair price. Buyers, banks, and insurers stay out for the same reason: the asset is invisible. Nobody can answer the three questions that underwriting requires — where is it, how healthy is it, and is it really there.

Traditional livestock-tracking solutions were designed for European farms with constant power, constant LTE, and a vet visit every quarter. None of that applies in the Sahel. Sensors die, data never syncs, and the last-mile herder never sees a working product. The result is a market stuck on paper: agricultural lending in the region runs at single-digit penetration while cattle insurance is almost non-existent.

We think the unlock is not a better tag or a better app. It is a trusted third-party that sits between the herd and the financial institution, producing a verifiable record both sides can act on.

The wedge is Edge AI. Running computer vision on a solar-powered node, right next to the herd, solves the three constraints that defeated previous attempts. It removes the dependency on permanent connectivity — the node keeps observing and logging even when there is no network for days. It drops the energy budget low enough for a small panel to keep it alive through the dry season. And it puts inference where the data is born, so we only sync what matters: events, not raw video.

YOLOv4-Tiny on a Raspberry Pi class device is now cheap enough, fast enough, and accurate enough to detect behavioural signals — feeding, rumination, lameness, distress — that correlate with health and loss events. Five years ago this pipeline required a workstation and a cloud round-trip. Today it runs on a €70 board that a herder's cooperative can afford to rent.

Three layers. The edge layer is the solar node itself: a Raspberry Pi (and Orange Pi RK3588 for higher-density sites) running YOLOv4-Tiny, a local MQTT broker, and offline-first storage. Events queue locally and ship opportunistically over any available channel — 4G, 2G, LoRa, or a human walking past a mesh relay.

The cloud layer is a Python / Django backend on GCP with PostgreSQL for structured records and InfluxDB for telemetry time-series. This is where herd health, provenance, and behavioural timelines become a queryable truth source. Dashboards on top, APIs beneath.

The partner layer is the reason the first two exist. We expose a signed verification API to banks, microfinance institutions, and insurers — they query DomoSo to confirm a herd exists, is healthy, and has behaved normally over a given window, and they use that answer to price a loan or an insurance policy. We are the trusted third-party; we do not lend and we do not insure.