Portuguese 
Spanish 
Analysis With Artificial Intelligence Reveals Hidden Celestial Objects in NASA’s Public Data, Using Advanced Methods and Astronomical Databases Accumulated Since 2009
Since 2009, NASA’s Wide-field Infrared Survey Explorer (WISE) mission has been mapping the sky in infrared. However, it wasn’t until years later that this data gained new relevance. In 2025, American student Matteo Paz, then 18, applied artificial intelligence to this collection and identified over 1.5 million previously unknown cosmic objects.
Although the records had been available for years, the extreme volume of information made detailed analysis challenging. Therefore, the application of machine learning algorithms was crucial. As a result, the research demonstrated how technology can reinterpret raw data and generate relevant scientific discoveries.
Consequently, the study was published in The Astronomical Journal in 2025. Moreover, the work received recognition in one of the most prestigious scientific student competitions in the United States, reinforcing its academic credibility.
-
Pennsylvania Mine Creates Ice in Summer Heat and Melts in Winter, Turning a Mountain Crevice into a Reverse Natural Refrigerator
-
China Accelerates Global Expansion of Robotics: From Smart Vacuums to Industrial and Humanoid Machines, Millions Exported to Over 150 Markets
-
Polar Air Mass to Hit Brazil: Temperatures to Plummet in Nine States, Near Freezing in the South with Frost Risk Until Monday
-
Chinese “dark factories” produce electric cars 24/7 with minimal human intervention, as Zeekr assembles 800 units daily, raising concerns in Brazil.
NASA Mission Accumulated Billions of Data Over the Years
Since the start of the WISE mission in December 2009, the entire sky has been observed in search of sources of infrared radiation. As a result, the project accumulated about 200 billion records, known as “appearances”. These data document variations in brightness and energy emission over time.
However, due to the scale of this database, many discoveries remained hidden. Thus, the main limitation was not data collection, but the absence of tools capable of processing it efficiently.
Young Researcher Adopted Variability-Based Approach
Given this scenario, Matteo Paz chose a different strategy. Instead of analyzing static images, he focused his research on temporal variations in brightness, typical of variable stars. This category includes unstable young stars, active black holes, and explosive events like supernovae.
However, manually processing billions of measurements would be unfeasible. Therefore, artificial intelligence became essential for automating the analysis and accelerating the identification of relevant patterns.
VARnet Algorithm Expanded Astronomical Analysis Capabilities
To overcome the technical challenge, Matteo developed VARnet, an algorithm capable of analyzing astronomical time series at high speed. The system combines deep learning, convolutional neural networks, wavelet decomposition, and Fourier transforms, allowing for the recognition of nearly imperceptible patterns.
Moreover, each source could be analyzed in less than a millisecond, with support from modern GPUs. After training with over 1 million simulated light curves, the model achieved high accuracy. As a consequence, over 1.5 million candidates for variable objects, including unprecedented sources, were identified.
The development was supported by the Infrared Processing and Analysis Center (IPAC), linked to the California Institute of Technology (Caltech), ensuring scientific validation compatible with professional standards.
Scientific Recognition and Impact for Future Research
In 2025, Matteo Paz’s work earned him the top prize at the Regeneron Science Talent Search, worth US$ 250,000. In addition to the academic recognition, the study highlighted a structural change in the exploration of large astronomical databases.
Finally, the VARnet catalog was made publicly available at the end of 2025. Since then, it has served as a basis for new studies, including projects such as the Vera C. Rubin Observatory, which is expected to conduct the largest astronomical survey in history, reinforcing that new discoveries can arise from new readings of already observed data.
