Skip to main content
Back to challenges
Challenge

Build an El Niño Agricultural Risk Forecaster with Google ADK and DeepSeek R1

The global weather pattern is undergoing drastic shifts driven by El Niño, leading to severe droughts in Southeast Asia and South America while disrupting agricultural supply chains. Insurance underwriters urgently require predictive multi-modal systems capable of combining climate indices, sea surface temperature (SST) anomalies, and satellite crop health signals to forecast crop loss risk prior to harvest season. In this challenge, you will implement an autonomous climate risk scoring agent using Google ADK (Agent Development Kit) integrated with DeepSeek R1 for reasoning over complex meteorological telemetry. You will ingest netCDF weather grids, compute oceanic Niño indices, and process multispectral Sentinel imagery to compute localized crop vulnerability indices. Your solution will emit calibrated risk scores and structured actuarial notes for agricultural parametric insurance policies across vulnerable regions.

Machine LearningHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Build an agentic agricultural climate risk system using Google ADK and DeepSeek R1 to ingest weather grids and assess regional crop failure risk.

Delivery guide

How work is evaluated

Evaluation

Evaluates agent ability to calculate drought risk scores from synthetic netCDF climate grids and produce structured risk evaluations.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

schema_validation

Output adheres strictly to expected JSON structure

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

risk_score_mae

Mean Absolute Error against target actuarial baseline • target: 0.05 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Build a multi-modal agent workflow with Google ADK and DeepSeek R1.

  • Process spatial-temporal NetCDF climate data and satellite SST anomalies.

  • Generate calibrated parametric insurance trigger recommendations with uncertainty bounds.

  • Evaluate agricultural drought resilience across Southeast Asia and Latin America.

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic time-series and spatial weather grids simulating El Niño conditions across rice and wheat growing regions, including SST anomalies, rainfall deficits, and historical claims.

Challenge setup

Evaluates agent ability to calculate drought risk scores from synthetic netCDF climate grids and produce structured risk evaluations.

Challenge input

JSON containing lat/lon bounds, month, and SST array

Google ADK

Official framework required for building multi-turn agentic workflows.

DeepSeek R1

Provides chain-of-thought mathematical and climate logical reasoning.

Evaluated output

JSON containing risk_score (0-1), trigger_event (bool), and confidence_interval

Evaluation checks
  • Output adheres strictly to expected JSON structure
  • Mean Absolute Error against target actuarial baseline • target: 0.05 • range: 0-1
Configured evaluation evidence
  • Benchmark: ClimateRiskBench-Agri
  • Risk Score Mae target: 0.05
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • DeepSeek · Required
  • Google · Optional
  • Google ADK · Optional

Evaluation contract

  • schema_validation · Weight 1
  • risk_score_mae · Weight 1

Metrics

  • risk_score_mae · Target 0.05

Benchmarks

  • ClimateRiskBench-Agri
  • energy

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing lat/lon bounds, month, and SST array

    Output: JSON containing risk_score (0-1), trigger_event (bool), and confidence_interval

    Execution harness: Python

Evaluation tools

  • Google ADK · Agent System
  • DeepSeek R1 · Reasoning Model

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Build an El Niño Agricultural Risk Forecaster with Google ADK and DeepSeek R1