NASA forecasting challenge: solar-wind prediction

[Enrico Camporeale]

Dear colleagues, 

We are excited to announce the Benchlab Solar Wind Prediction AI Tournament, a real-world forecasting and operational-ML challenge evaluated on live, future solar-wind data.

Unlike traditional ML competitions, Benchlab does not evaluate models on a static test set. Submitted workflows run in a forward-time operational loop, forecasting ambient solar-wind speed 72 hours ahead using only information available at the time of inference.

No prior space-weather background is required. We are actively seeking students, early-career researchers, data scientists, ML engineers and scientific software developers with skills in time-series forecasting, computer vision, scientific Python, containerization and robust ML pipeline design.

The challenge:
- Goal: predict ambient solar-wind speed 72 hours in advance.
- Submission: package code in a Docker container matching the provided template.
- Live evaluation: submitted workflows must run statelessly, fetch their own data dependencies at inference
time, and handle missing data or latency without crashing.
- Recognition: participants will compete for a $20,000 overall cash prize fund, with additional wildcard categories to be defined.
Individuals and interdisciplinary teams of up to four members are welcome.

Register your interest here:
https://fdl.ai/benchlab-solar-wind-prediction-2026

Bring your data-science expertise to a high-impact forecasting problem with relevance for satellites, communications, navigation, power grids and space operations.

Best regards,
Enrico Camporeale and Nina Bonaventura