Imagine scrolling through billions of infrared snapshots of the night sky and spotting a tiny, almost‑imperceptible flicker that could be a distant asteroid, a dying star, or a hidden exoplanet. That’s exactly what a California teenager did—using a homemade artificial‑intelligence model to sift through NASA’s massive NEOWISE archive and flag more than 1.5 million variable sources. This breakthrough isn’t just a cool story; it’s a glimpse of how the next generation will explore our solar system and the wider cosmos.
⚡ Quick Answer
Key point: A high‑schooler’s DIY AI uncovered 1.5 million faint, time‑varying objects in NASA’s infrared sky archive, demonstrating that powerful data mining tools are now within reach of anyone with curiosity and a laptop.
🌌 The Treasure Trove: NASA’s NEOWISE Archive
Since 2013, the NEOWISE (Near‑Earth Object Wide‑field Infrared Survey Explorer) spacecraft has been circling Earth, repeatedly scanning the entire sky in infrared wavelengths. Over a decade, it has amassed petabytes of data—essentially a time‑lapse movie of the infrared universe, capturing everything from bright comets to faint distant galaxies.
Because infrared light can penetrate dust clouds that hide visible‑light objects, NEOWISE is uniquely suited for detecting cool bodies like asteroids, brown dwarfs, and even the subtle heat signatures of variable stars. However, the sheer volume of images makes manual inspection impossible; traditional pipelines focus on known targets, leaving countless faint variations unexamined.
📌 Key Facts:
- 🔴 Mission start: NEOWISE was repurposed in 2013 after its original WISE mission finished.
- 🌡️ Wavelength range: 3.4 µm and 4.6 µm infrared bands, ideal for spotting cool, dark objects.
- ⏱️ Data volume: Over 2 million individual sky frames, each covering 47 square degrees.
🤖 From Kitchen Table to Cosmic Detective: The Teen’s Homemade AI
Inspired by open‑source machine‑learning libraries, the teenager—named Alex Paz—built a lightweight neural network on a personal laptop. The model combines two classic signal‑processing tools: Fourier transforms, which decompose a signal into its constituent frequencies, and wavelet analysis, which captures localized, transient changes.
By feeding the AI sequences of brightness measurements for each sky pixel over time, the system learned to distinguish genuine astrophysical variability from instrumental noise. The result? A catalog of 1.5 million objects that brighten, dim, or blink on timescales ranging from minutes to months.
🪐 Why Variable Objects Matter for Planetary Science
Variable sources are the cosmic breadcrumbs that lead us to hidden phenomena:
- Asteroids & comets: Small changes in infrared brightness can signal rotation, shape irregularities, or surface activity—key clues for impact‑risk assessment and resource utilization.
- Trans‑Neptunian objects: Faint, slow‑varying light curves help identify distant icy bodies that shape our understanding of the solar system’s formation.
- Variable stars: Pulsating or eruptive stars serve as distance markers (standard candles) and laboratories for stellar physics, which indirectly inform planetary formation models.
Detecting such subtle signals expands the inventory of objects we can study, refining orbital models, composition estimates, and even climate histories of distant worlds.
💫 Democratizing Big Data in Astronomy
Until recently, analyzing terabytes of sky data required supercomputers and dedicated research teams. Paz’s project shows that with open‑source tools, modest hardware, and a solid grasp of mathematics, anyone can contribute to frontier science.
This shift mirrors the broader “citizen‑science” movement—think of projects like Zooniverse—but now the contribution can be algorithmic rather than manual, allowing for faster, deeper discoveries.
🚀 The Rubin Observatory Connection
In February 2026, the Vera C. Rubin Observatory in Chile turned on its real‑time alert stream, broadcasting roughly 800,000 transient alerts in a single night. This flood of data is the next logical step for AI‑driven pipelines like Paz’s.
By integrating his model with Rubin’s alert system, astronomers can automatically flag objects that exhibit the same subtle infrared variability, cross‑matching them with optical detections to refine classifications. This synergy accelerates follow‑up observations with telescopes worldwide, from the James Webb Space Telescope to ground‑based spectrographs.
🔭 How Educators Can Bring This Into the Classroom
Teachers can turn this story into hands‑on lessons:
- Data mining basics: Use free Python notebooks to download a small subset of NEOWISE data and apply Fourier transforms.
- AI building blocks: Introduce simple neural‑network libraries (e.g., TensorFlow Lite) to classify light curves.
- Cross‑disciplinary links: Connect variability to planetary rotation, surface composition, and even climate cycles on exoplanets.
Such projects empower students to see themselves as contributors, not just consumers, of scientific knowledge.
🎯 Key Takeaways
- ✨ Point 1: A California teenager’s DIY AI identified 1.5 million variable objects in NASA’s NEOWISE infrared archive.
- ✨ Point 2: The approach uses Fourier and wavelet analysis to reveal faint, time‑dependent signals that traditional pipelines miss.
- ✨ Point 3: This work exemplifies the democratization of big‑data astronomy and will complement upcoming massive alert streams from the Rubin Observatory.