About

We're building an insurance backbone for the world's smallholder farmers.

There are roughly 500 million smallholder farmers worldwide. Most of them are locked out of the crop-insurance system that already exists — not because they don't qualify, but because the paperwork is in the wrong language, demands evidence they can't structure, and assumes literacy in domains they were never trained for. Less than 20% of eligible farmers ever file a claim after a loss.

The agent stack we needed to fix this just arrived. Multimodal vision models that can identify crops and assess damage from a phone photo. Multilingual reasoning models that can talk to a farmer in Bengali or Swahili and reply in the same language. Vector stores that ground decisions in the agronomy literature and surface fraud patterns. ClaimFarm is what happens when you wire those pieces together with a human adjuster in the loop.

Our north star: a farmer in rural India loses a third of her wheat to late-season hail, sends one photo over WhatsApp, and 24 hours later money lands in her bank account. No forms. No language barrier. No $10 lost on a bus ride to the insurance office.

Principles

How we build.

Humans in the loop, always
Every approval has a human adjuster signing off. The model surfaces evidence; the human owns the decision.
Build for the worst phone
WhatsApp / Telegram first. We assume 2G, low literacy, no app installs.
Show your work
Every AI verdict comes with structured evidence (visible indicators, weather signals, retrieval context). No black boxes.
Privacy by default
Personal photos and IDs are encrypted at rest. Signed URLs only. GDPR-ready data deletion.