Real-time Weather Prediction Agents
This challenge requires you to build a multi-agent system using CrewAI for real-time weather forecasting. Your system will simulate processing continuous streams of satellite telemetry and sensor data via Apache Kafka. A team of specialized agents, powered by Gemini 2.5 Pro for data analysis and reasoning, will collaborate to perform data ingestion, pattern recognition, predictive modeling, and alert generation. The focus is on orchestrating complex, asynchronous workflows where agents manage data integrity, learn from historical patterns, and provide actionable weather insights with low latency.
What you are building
The core problem, expected build, and operating context for this challenge.
Orchestrate a multi-agent system with CrewAI and Gemini 2.5 Pro for real-time weather forecasting from simulated satellite data streams.
How work is evaluated
Evaluation will focus on the accuracy and timeliness of weather forecasts generated by the agent team, the robustness of the Kafka integration, and the collaborative efficiency of the CrewAI agents.
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.
KafkaDataIngestion
Verify that the agent system successfully consumes data from the specified Kafka topic.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
ForecastCompleteness
Ensure a 24-hour forecast is generated without missing data points.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
ForecastAccuracy (RMSE)
Root Mean Square Error for temperature predictions compared to ground truth, lower is better. • target: 1.5 • range: 0-5
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
ProcessingLatency
Average time taken from data ingestion to forecast output, lower is better. • target: 200 • range: 0-500
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Master CrewAI for defining roles, goals, tasks, and tools for collaborative AI agents.
Implement real-time data ingestion pipelines using Apache Kafka for high-throughput sensor data.
Design a 'Data Analyst' agent leveraging Gemini 2.5 Pro's capabilities for complex pattern recognition in meteorological data.
Develop a 'Predictive Modeler' agent responsible for running forecasting models and updating predictions based on new data.
Integrate Weights & Biases (W&B) for tracking model experiments, managing different forecasting models, and monitoring their performance.
Deploy a simulated inference endpoint using Cerebras Inference for serving weather prediction models with low latency.
Orchestrate an end-to-end workflow where agents automatically detect anomalies, generate forecasts, and publish alerts.
Reference links and supporting material
Sample data for 1 tasks
Challenge setup
Evaluation will focus on the accuracy and timeliness of weather forecasts generated by the agent team, the robustness of the Kafka integration, and the collaborative efficiency of the CrewAI agents.
Challenge input
{'kafka_topic': 'string', 'duration_seconds': 'integer', 'ground_truth_forecast': 'list_of_dicts'}
Agent execution
No agent tool is listed for this challenge.
Evaluated output
{'generated_forecast': 'list_of_dicts', 'processing_latency_ms': 'float', 'alerts_issued': 'list_of_strings'}
- Verify that the agent system successfully consumes data from the specified Kafka topic.
- Ensure a 24-hour forecast is generated without missing data points.
- Root Mean Square Error for temperature predictions compared to ground truth, lower is better. • target: 1.5...
- ForecastAccuracy (RMSE) target: 1.5
- 1 public reference case
- Python execution harness
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
No tool records are attached.
Evaluation contract
- KafkaDataIngestion · Weight 1
- ForecastCompleteness · Weight 1
- ForecastAccuracy (RMSE) · Weight 1
- ProcessingLatency · Weight 1
Metrics
- ForecastAccuracy (RMSE) · Target 1.5
- ProcessingLatency · Target 200
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: {'kafka_topic': 'string', 'duration_seconds': 'integer', 'ground_truth_forecast': 'list_of_dicts'}
Output: {'generated_forecast': 'list_of_dicts', 'processing_latency_ms': 'float', 'alerts_issued': 'list_of_strings'}
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.