1. Upload light-curve table
CSV with one row per observation. Required headers shown in the schema.
A World Away: Intelligent Exoplanet Hunter
Our Argonautas crew blends NASA light-curve archives with explainable AI to surface exoplanet candidates fast, then delivers stories and visuals that keep scientists and citizen astronomers in sync.
Exoplanet AI toolkit
Download the production XGBoost light-curve classifier and run it locally or in your own cloud notebook. Use the interface below to run a batch upload or a single-candidate review.
Choose a model
exoplanet_xgb_pipeline.joblib
Trained on confirmed Kepler & TESS transits - 7 engineered features
kepler_disposition_model.joblib
Disposition model tuned on Kepler vetting labels - CSV batches only
2. Evaluate a single candidate
Adjust the controls to shape a candidate and estimate its likelihood.
The Kepler disposition model only supports CSV uploads. Switch back to Exoplanet XGBoost to use the interactive sliders.
Switch chart tabs or click data points to focus a feature.
Runs the values through exoplanet_xgb_pipeline.joblib on the server.
Exoplanets
In 1584, an Italian philosopher named Giordano Bruno dared to conjecture that, beyond the planets known to Roman tradition (Mercury, Venus, Mars, Jupiter, and Saturn), there existed infinite planets around the infinite points that filled the celestial sphere.
Despite the misfortune and censorship that his contrasting ideas provoked, they endured in the scientific imagination for centuries until his long-denied hypothesis was confirmed in 1992 by Aleksander Wolszczan and Dale Frail. With the discovery of the first exoplanet—the name given to planetary bodies beyond the Solar System—our view of the universe changed radically and a new branch of science opened: exoplanetary astronomy. Today we know of more than 5,000 exoplanets, an achievement that required machines in space and human ingenuity on Earth.
However, just as there are many discovered exoplanets, an infinity still waits to be found. The vast amounts of data provided by telescopes make the task of analyzing every potential hint of a planet orbiting a star other than the Sun arduous, if not unattainable.
With new technological revolutions and the immense development of Artificial Intelligence, automated tasks lighten researchers' workload—and, more importantly, they enable discoveries that would be nearly impossible through human effort alone.
How do you discover an exoplanet?
A light bulb, under normal conditions, should stay on without difficulty. If we see the bulb dim, it may have been caused by an electrical issue or because something blocked the bulb from our point of view. If the bulb dims again, there might already be a structural problem. If it happens yet again, it is time to analyze what is going on.
For stars, the most common scenario is that they can maintain their brightness for billions of years—a span that is practically infinite on human timescales. Therefore, when scientists observe a star dimming, it becomes a subject of serious study. A star can vary its brightness because of its own conditions, because it shares an orbit with another star, or—most importantly—because a planet is orbiting it.
When an exoplanet passes in front of a star, the same thing happens as when a fly or object temporarily blocks a light bulb: the brightness decreases. This drop, known scientifically as a light curve, allows scientists to tell whether an exoplanet is present, and it even reveals how large the planet is and how it orbits its star.
What are we doing?
Within the Argonautas initiative, we harness the best of machine learning and astronomy to facilitate the study of confirmed, candidate, and rejected exoplanets, while also making the process of scientific discovery more accessible to everyone.
