NASA’s AI-Powered Hunt Reveals Thousands of Hidden Exoplanets

The search for worlds beyond our solar system has entered a bold new era, thanks to artificial intelligence and open science. Scientists have now identified over 6,000 exoplanets—planets that orbit stars other than our Sun—yet many more remain hidden within NASA’s treasure trove of data. From the retired Kepler mission to the ongoing TESS (Transiting Exoplanet Survey Satellite) mission, these space telescopes have generated an unprecedented amount of publicly available data, providing researchers worldwide with the raw materials to uncover new planets. But the sheer scale of these datasets presents a challenge: how do you sift through hundreds of thousands of signals to find genuine exoplanets? The answer increasingly lies in AI.

In 2021, NASA’s Ames Research Center team developed ExoMiner, an open-source artificial intelligence tool that validated 370 new exoplanets from Kepler data. Building on that success, the team has now unveiled ExoMiner++, an upgraded AI model trained on both Kepler and TESS datasets. In its initial run, ExoMiner++ flagged 7,000 potential exoplanet candidates from TESS data—signals likely caused by planets but requiring further observation to confirm.

The software works by analyzing possible transit signals—tiny dips in a star’s brightness caused by a planet passing in front of it—and distinguishing them from false positives, such as eclipsing binary stars. “When you have hundreds of thousands of signals, this is the ideal place to deploy deep learning technologies,” said Miguel Martinho, co-investigator of ExoMiner++. By combining the wide-sky survey approach of TESS with Kepler’s deep, targeted observations, ExoMiner++ can leverage the strengths of both missions to deliver highly accurate results.

ExoMiner++ is freely available on GitHub, allowing researchers everywhere to use the tool to explore TESS’s growing public data archive. Open-source tools like this accelerate scientific discovery, says NASA Chief Science Data Officer Kevin Murphy, by allowing replication of results and deeper dives into the data. Future versions of ExoMiner++ aim to not only flag candidate signals but also identify them directly from raw telescope data, further streamlining the hunt for new worlds.

NASA’s commitment to open science ensures that data from future missions, including the Nancy Grace Roman Space Telescope, will also be publicly available, offering ExoMiner users tens of thousands more potential exoplanet transits to analyze. This combination of open-source software, AI, and freely shared data is driving the rapid growth of exoplanet science.

What Undercode Say:

ExoMiner++ represents a remarkable synergy between AI and astronomy, showcasing how machine learning can accelerate discoveries that would take human researchers decades to achieve alone. By automating the validation of exoplanet candidates, the software not only saves enormous time but also improves accuracy, reducing human error in the face of vast datasets.

The public availability of both the data and the tool underscores a broader trend in modern science: open-source approaches are not just a philosophical choice but a practical one. With ExoMiner++, any research team, whether from a major university or an independent astronomer, can contribute to the growing catalog of exoplanets. This democratization of science enhances innovation, as diverse teams can apply novel techniques or explore overlooked signals.

ExoMiner++’s integration of Kepler and TESS data is particularly strategic. Kepler’s deep survey of a small sky patch provides high-resolution data ideal for training AI models, while TESS’s wide-field coverage ensures the model can generalize across a broader range of stars. This combination strengthens the reliability of ExoMiner++’s predictions.

Furthermore, the model’s design to identify transits from raw data in future versions suggests a shift toward more autonomous AI-driven astronomy. By detecting planets directly without human preprocessing, ExoMiner++ could exponentially increase the pace of discovery.

The long-term implications extend to upcoming missions like the Nancy Grace Roman Space Telescope. Its unprecedented data volume will likely overwhelm traditional analysis methods. AI models such as ExoMiner++ will be indispensable, not just for discovery, but for understanding the demographics of exoplanets across the galaxy.

ExoMiner++ also exemplifies how open-source software can foster cross-disciplinary collaboration. Machine learning experts, astronomers, and citizen scientists can all leverage the same tools and datasets, creating a global network of discovery. This collaborative model could serve as a template for other scientific fields, from climate modeling to genomics.

Ultimately, ExoMiner++ is more than a software tool; it’s a proof-of-concept for AI-driven exploration in space. It reflects a future in which the combined power of human curiosity, advanced computation, and shared knowledge accelerates our understanding of the universe at an unprecedented pace.

Fact Checker Results:

✅ Data Volume Verified: NASA has confirmed over 6,000 known exoplanets discovered primarily via Kepler and TESS missions.
✅ ExoMiner++ Accuracy: Peer-reviewed Astronomical Journal paper confirms initial run identified 7,000 potential exoplanet candidates.
✅ Open-Source Availability: ExoMiner++ is publicly downloadable on GitHub for global research use.

Prediction:

🌌 The integration of AI like ExoMiner++ with future missions such as the Roman Space Telescope could triple or quadruple the rate of exoplanet discovery in the next decade.
🤖 Open-source AI tools may become standard in astronomy, making citizen scientists key contributors to major discoveries.
🚀 The next breakthroughs in identifying Earth-like planets are likely to emerge from AI-driven analysis of publicly available datasets, revolutionizing our understanding of potentially habitable worlds.