An AI Decision Copilot to Help India Face the 2026 El Niño Crisis

Built during the Google Cloud Gen AI Academy APAC Hackathon, the El Niño 2026 Decision Copilot is an explainable AI platform designed to help India navigate water, agriculture and food-security risks ahead of a strong El Niño year. It unifies scattered weather, satellite, reservoir and agricultural data into BigQuery as a single source of truth, then feeds an interpretable BigQuery ML model that produces district-level risk scores instead of relying on a black-box neural network.

At the core of the architecture are specialized AI agents built with Google's Agent Development Kit: a triage agent that ranks districts needing urgent attention, a deterministic allocation agent that decides how resources like water tankers and relief supplies should be distributed, and a RAG-powered farmer advisory agent that grounds crop recommendations in official contingency documents. Every output is paired with the data and citations behind it, so administrators and farmers never have to trust the AI blindly.

While the hackathon prototype covered only 23 high-risk districts, the platform has since scaled to cover 763 districts across India — a challenge that proved harder than the AI itself, since public datasets varied wildly in schema and quality. The author's key takeaway is that large language models become far more valuable as a reasoning layer over trustworthy, well-structured data rather than as the entire application, and that for public-sector decisions, explainability often matters more than model sophistication.