Parametric Drought Insurance Trigger Verification with AutoGen and Fireworks AI
Parametric insurance products rely on objective, automated weather indicators to payout claims rapidly without waiting for manual damage appraisal. During intense El Niño cycles, insurance providers receive millions of claims across decentralized rural micro-grids. An automated multi-agent consensus system is needed to verify satellite telemetry and prevent fraudulent or faulty parametric triggers. In this challenge, you will orchestrate a multi-agent verification system using Microsoft AutoGen powered by Fireworks AI inference acceleration. Agents will represent distinct domain functions: a Satellite Telemetry Verification Agent, an Actuarial Fraud Inspector Agent, and an Executive Settlement Agent. Your solution will execute automated peer conversations to ingest microclimate station data, detect sensor anomalies, compare historical El Niño indices, and issue cryptographically verifiable trigger settlements.
What you are building
The core problem, expected build, and operating context for this challenge.
Orchestrate an AutoGen multi-agent system powered by Fireworks AI to evaluate satellite weather telemetry and verify parametric insurance claim triggers.
How work is evaluated
Verifies that the multi-agent system correctly flags corrupted telemetry and authorizes legitimate parametric payouts.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
agent_consensus_reached
All agents reach consensus without infinite loop
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
fraud_detection_accuracy
Accuracy in rejecting faulty sensor feeds • target: 0.95 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Design a multi-agent conversational system using Microsoft AutoGen.
Deploy rapid LLM inference pipelines via Fireworks AI.
Implement multi-agent consensus algorithms for fraud and anomaly checking in sensor feeds.
Automate verifiable parametric payout calculations for climate disaster insurance.
Reference links and supporting material
Synthetic array of historical weather station readings during major El Niño events alongside simulated sensor corruption noise.
Challenge setup
Verifies that the multi-agent system correctly flags corrupted telemetry and authorizes legitimate parametric payouts.
Challenge input
JSON with policy_id, rainfall_mm_array, sensor_health_flag
AutoGen
Mandatory framework for multi-agent conversational orchestration.
Fireworks AI
Provides fast serverless inference for LLM agents.
Evaluated output
JSON with verified (bool), payout_amount, consensus_reasoning
- All agents reach consensus without infinite loop
- Accuracy in rejecting faulty sensor feeds • target: 0.95 • range: 0-1
- Benchmark: ParametricClaimsBench
- Fraud Detection Accuracy target: 0.95
- 1 public reference case
- 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
- AutoGen · Optional
- Zed · Optional
- Fireworks AI · Required
- Fireworks AI · Required
Evaluation contract
- agent_consensus_reached · Weight 1
- fraud_detection_accuracy · Weight 1
Metrics
- fraud_detection_accuracy · Target 0.95
Benchmarks
- ParametricClaimsBench
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON with policy_id, rainfall_mm_array, sensor_health_flag
Output: JSON with verified (bool), payout_amount, consensus_reasoning
Execution harness: Python
Evaluation tools
- AutoGen · Agent System
- Fireworks AI · Inference Engine
Recipe freeze marker
No recipe freeze time is recorded.