Simulate European Urban Heat Parametric Insurance with AutoGen and Together AI
Chronic heatwaves in European cities increase municipal public health expenditures and infrastructure downtime. Build a multi-agent simulation using AutoGen and Together AI to automate parametric payout triggers based on weather feed spikes. Achieve accurate automated payout triggers with zero false positives on baseline historical temperatures.
What you are building
Develop an AutoGen multi-agent system powered by Together AI fast inference to evaluate daily heat index metrics against parametric trigger thresholds.
How work is evaluated
Evaluates AutoGen agents on triggering parametric policy payouts under various heat event scenarios.
Datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
zero_false_positives
No parametric triggers fired during non-extreme weather periods.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
trigger_f1_score
F1 score of triggering payouts when actual thresholds are exceeded • target: 0.95 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Learning goals
Build a multi-agent collaborative workflow using AutoGen framework.
Leverage Together AI endpoint for low-latency time-series evaluation.
Construct parametric trigger verification logic for urban heat indices.
Design an auditable decision logging module for reinsurance settlement.
Resources
Historical and projected ERA5 European heatwave climate data formatted for parametric trigger benchmarking.
Challenge setup
Evaluates AutoGen agents on triggering parametric policy payouts under various heat event scenarios.
Challenge input
JSON list of daily weather observations (temp, humidity, duration)
AutoGen
Enables multi-agent conversational consensus.
Together AI
Provides fast open-weights LLM execution.
Evaluated output
JSON detailing triggered events, calculation proof, and total payout
- No parametric triggers fired during non-extreme weather periods.
- F1 score of triggering payouts when actual thresholds are exceeded • target: 0.95 • range: 0-1
- Benchmark: ParametricTriggerBench
- Trigger F1 Score target: 0.95
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Shows the first task, up to three tools, and up to three evidence items.
View technical recipe
Configured tools
- RAI · Optional
- 01.AI · Optional
- Together AI · Required
- Together AI · Required
Evaluation contract
- zero_false_positives · Weight 1
- trigger_f1_score · Weight 1
Metrics
- trigger_f1_score · Target 0.95
Benchmarks
- ParametricTriggerBench
- energy
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON list of daily weather observations (temp, humidity, duration)
Output: JSON detailing triggered events, calculation proof, and total payout
Execution harness: Python
Evaluation tools
- AutoGen · Multi-agent Framework
- Together AI · Inference Host
Recipe freeze marker
No recipe freeze time is recorded.