Track 4 finalist · Global AI Hackathon × Qwen Cloud

Insurance claims for smallholder farmers, filed in 60 seconds.

ClaimFarm turns a single WhatsApp or Telegram photo into a fully filed crop-insurance claim: multimodal AI damage assessment, weather corroboration, fraud detection, multilingual replies, human adjuster review. Built for the 500 million farmers locked out by paperwork.

500M
smallholder farmers globally
< 20%
ever file an insurance claim
10+
languages supported on day one
60s
from photo to filed claim

How it works

Photo in. Filed claim out.

01

Farmer sends a photo

From WhatsApp or Telegram. Plus a short caption in their own language. No app to install.

02

AI assesses damage

Qwen-VL identifies the crop and damage. Open-Meteo corroborates against the weather record.

03

Adjuster reviews

ClaimFarm's console surfaces the AI verdict, similar past claims, and any fraud flags side-by-side.

04

Insurer receives a clean claim

A pre-filled PDF claim, evidence-stamped, lands in the insurer's intake, ready to pay out.

What's inside

Six AI checks before a single dollar moves.

Multimodal damage assessment

Qwen-VL-Max inspects the photo: identifies crop, classifies the damage (drought, flood, hail, pest, disease, frost, wind), scores severity and affected area, and returns structured evidence.

Weather corroboration

Open-Meteo historical data is pulled for the farm's GPS and the photo's capture date; Qwen-Max judges whether the weather supports the visual verdict.

Retrieval + fraud detection

Qwen embeddings + Alibaba DashVector surface similar past claims and flag suspicious narrative reuse. Same-farmer near-duplicates are blocked at >0.93 cosine similarity.

Photo forensics

EXIF capture time, GPS, camera make, edit-software fingerprint, plus a Qwen-VL authenticity check that flags watermarks, screenshots, and AI generations.

Multilingual replies

Detects the farmer's language and replies in it. Hindi, Bengali, Swahili, Spanish, Portuguese, French, Arabic, Indonesian, Chinese, English on launch.

Human-in-the-loop review

Every claim lands in an adjuster console with the AI verdict, evidence, similar past claims, and risk flags. One click submits to the insurer.

Built on

Qwen Cloud + Alibaba Cloud, end to end.

Qwen-VL-Max
Multimodal damage assessment, photo authenticity verdict.
Qwen-Max
Weather corroboration, claim drafting, multilingual replies.
text-embedding-v3
RAG over past claims and agronomy KB; fraud similarity.
Alibaba DashVector
1024-dim cosine vector store for past claims + agronomy.
Alibaba OSS
Encrypted photo + PDF storage with signed URLs.
Function Compute 3.0
Serverless custom-container backend, public HTTPS trigger.
Open-Meteo
Historical weather archive for damage corroboration.
WeasyPrint
Polished PDF claim packets for insurer intake.

Stop losing claims to paperwork.

Smallholder farmers don't need an app. They have WhatsApp. Plug ClaimFarm in and your insurer intake speaks photo.