🤓 Latest Submissions
AFYARISK-ENGINE
PROBLEM Health insurance in Africa faces a critical data infrastructure gap. Insurers across East Africa operate five disconnected workflows using outdated tools: underwriters price premiums using static age-based tables with no clinical data, fraud analysts review claims manually catching less than 30% of fraudulent submissions, actuaries calculate IBNR reserves quarterly in Excel spreadsheets, agents answer policy questions by searching PDFs, and pricing models run independently from real-time risk data. The consequences are severe: loss ratios exceeding 80%, billions lost annually to undetected fraud, delayed reserve adjustments that breach regulatory requirements, and slow claims experiences that erode policyholder trust. Smaller insurers lack the engineering resources to build ML infrastructure, leaving them permanently behind. SOLUTION AfyaRisk 2.0 is a unified actuarial AI engine that replaces all five siloed workflows with a single real-time API platform. Built specifically for the African health insurance market, it delivers: 1. Clinical Risk Scoring — A Random Forest classifier takes patient biomarkers (glucose, BMI, age, blood pressure, insulin) and instantly classifies patients as Low, Medium, or High risk with survival probability and 3-year health degradation estimates. 2. Dynamic Premium Pricing — The risk score feeds directly into an actuarial pure premium engine using the formula: Premium = (Frequency × Severity) / (1 − Loss Ratio − Expense Loading), with risk-tier multipliers of 1.0×, 1.5×, and 2.5×. 3. Fraud Detection — An Isolation Forest anomaly detector scores every claim 0–100% for fraud probability, triggers explainable rule-based flags (HIGH_CLAIM_AMOUNT, RAPID_RESUBMISSION, EARLY_POLICY_LARGE_CLAIM), and returns APPROVE / REVIEW / REJECT recommendations. 4. IBNR Loss Reserving — The Chain-Ladder method computes Incurred But Not Reported reserves from historical claims development triangles — replacing manual actuarial spreadsheet work with a millisecond API call. 5. Policy RAG Search — A FAISS vector index powered by sentence-transformers allows underwriters and agents to search policy documents using plain English, returning semantically matched clauses with similarity scores. TARGET USERS: Health insurers, reinsurers, actuarial teams, and insurtech platforms across East Africa. The API-first architecture means any insurer's existing claims portal, mobile app, or core banking system can integrate all five engines through standard REST endpoints. The name Afya means health in Swahili reflecting the platform's African first mission.
27 Sep 2026
