Isomkevin/mizizi-kenya-ai

# Mizizi **Enterprise agricultural risk intelligence for Kenya** Mizizi is an enterprise agricultural risk intelligence platform. It helps lenders, cooperatives, and field officers see, verify, and explain agricultural credit risk using signals that already exist in the field — cooperative history, mobile money, input purchases, climate, and community relationships. **ZK Credit Rails** (Stellar Hacks: Real-World ZK): farmers prove credit tier from simulated mobile-money turnover and repayment history **without revealing raw transactions**. Groth16 proofs anchor credentials on Stellar testnet Soroban. See docs/stellar-hacks-zk-submission-plan.md. ## Built by **LESOM Dynamics** ## The problem Agricultural finance rarely fails for lack of capital. It fails because risk is hard to **see**, **verify**, and **explain**. Loan officers work with fragmented data. Farmers rarely understand why a decision was made. Mizizi connects those signals into one auditable view and keeps humans in the loop: AI explains; officers decide. --- ## What Mizizi does Four intelligence layers work together: | Layer | Purpose | | ------------------------ | ------------------------------------------------------------------------------------------------------------------ | | **Graph intelligence** | Entity relationships — farmers, cooperatives, loans, dealers, climate zones — for trust paths and risk propagation | | **Explainable AI** | Grounded recommendations with factor breakdowns, graph evidence, and officer override with audit trail | | **Climate intelligence** | County- and parcel-level weather signals tied to lending impact | | **Financial analytics** | Portfolio KPIs, repayment behaviour, geographic exposure, and decision quality metrics | Design principle …