CERES was built because the gap between when a famine becomes predictable and when the humanitarian system acts is measured in lives. Current early warning systems provide 30–45 days of actionable lead time. Pre-positioning food aid, mobilising logistics, and securing emergency funding requires 60–90 days.
CERES is an open, falsifiable, probabilistic forecasting system that synthesises six data streams into 90-day famine risk predictions — designed to give the humanitarian system the lead time it currently lacks.
It is free. It is open. Its methodology is published. Its predictions are timestamped and graded against outcomes. It is built to be scrutinised.
CERES is built on the Hypothesis Generation Engine (HGE) — AI-native infrastructure developed by Northflow Technologies for institutional-grade intelligence across complex, data-rich domains.
HGE is designed to do one thing: synthesise multi-source signals into ranked, evidenced, falsifiable hypotheses. It is not a dashboard. It is not a threshold alert system. It is a system that reads the evidence and tells you — with calibrated confidence — what it thinks is happening and why.
CERES is HGE Adapter #5. Each adapter applies the same hypothesis engine to a different domain. The same rigour. The same calibration standards. The same commitment to auditable reasoning.
CERES is a research and early warning tool. It is not an operational replacement for IPC field assessments, humanitarian programme decisions, or famine declarations.
Only the IPC Global Platform, through its established cadre process and field verification, has the mandate to declare famine (IPC Phase 5). A CERES Tier I prediction indicates a high probability of reaching IPC Phase 3+ — a signal for preparedness, not a declaration.
CERES predictions should be interpreted alongside, not instead of, FEWS NET situation reports, WFP VAM assessments, and field-based humanitarian intelligence.
Every design decision in CERES — from the calibrated confidence intervals to the publicly timestamped prediction ledger to the published limitations section — is made with institutional reviewers in mind.
The system is designed to be reviewed by WFP technical staff, FAO analysts, EU ECHO programme officers, academic food security researchers, and independent funders. It is built to withstand scrutiny, not to impress with outputs.
CERES is live as of 28 February 2026. Predictions are timestamped and graded against IPC outcomes at T+90 days. The model was initialised against 87 IPC transition records (2011–2023, 31 countries). Prospective forward validation is ongoing and publicly visible in the Validation Ledger. The full methodology is published at arXiv:2603.09425.
CERES is an open system seeking institutional partners for validation, deployment, and co-development. If you represent a humanitarian organisation, research institution, or funding body, we welcome your engagement.