AI for Smarter Telescopes: UK Universities' Groundbreaking Research

The Rise of AI in UK Astronomical Research

Artificial intelligence (AI) is transforming the field of astronomy by enabling telescopes to process vast amounts of data more efficiently, detect rare cosmic events, and operate autonomously. In the United Kingdom, universities and research institutions are at the forefront of this revolution, developing tools that make telescopes 'smarter' through real-time analysis, anomaly detection, and predictive maintenance. This integration of AI not only accelerates discoveries but also addresses the data explosion from next-generation observatories.

UK astronomers face petabytes of data daily from surveys like the Asteroid Terrestrial-impact Last Alert System (ATLAS) and upcoming facilities such as the Square Kilometre Array (SKA). Traditional methods struggle with noise—artefacts from instruments, atmosphere, or satellites—requiring manual sifting. AI steps in by learning patterns, classifying events, and prioritizing genuine signals, allowing scientists to focus on interpretation rather than filtering.

Oxford University's Virtual Research Assistant Transforms Supernovae Hunting

The University of Oxford has pioneered the Virtual Research Assistant (VRA), an AI system designed to sift through ATLAS alerts for supernovae—exploding stars beyond our galaxy. ATLAS, comprising five global telescopes, scans the visible sky every 24-48 hours, generating millions of potential transients nightly. After initial filters, 200-400 candidates remain, overwhelming human reviewers.

VRA employs decision-tree algorithms mimicking expert judgment, ranking alerts based on features like brightness changes and position. Trained on just 15,000 examples, it runs on standard laptops and integrates with the Lesedi Telescope in South Africa for automatic follow-ups. In its first year from December 2024, VRA cut alerts to humans by 85% while retaining over 99.9% of real supernovae, missing fewer than 0.08%.

• Step 1: Ingest ATLAS alerts and assess initial probability.
• Step 2: Update rankings over multiple nights as sky patches are reobserved.
• Step 3: Flag top candidates for robotic telescope confirmation.
• Benefits: Frees astronomers for science; scales to Vera C. Rubin Observatory's 10 million nightly alerts.

Lead researcher Dr. Héloïse Stevance, funded by the Schmidt AI in Science Fellowship, published findings in The Astrophysical Journal in September 2025. This tool exemplifies how Oxford is preparing for the Legacy Survey of Space and Time (LSST), launching in 2026.Read the Oxford VRA paper

Few-Shot AI Learning: Spotting Cosmic Events with Minimal Data

Building on VRA, Oxford physicists collaborated with Google Cloud on a few-shot learning model using Gemini large language model (LLM). Provided just 15 example images and instructions, it classifies real changes—like supernovae, variable stars, or asteroids—from artefacts with 93% accuracy, rising to 96.7% via self-correction.

A panel of 12 astronomers verified explanations as coherent, enabling trust in AI decisions. Published in Nature Astronomy (October 2025), this approach suits sparse training data in astronomy. Co-leads Dr. Fiorenzo Stoppa and Professor Stephen Smartt highlight its potential for dynamic sky surveys.Explore the Nature Astronomy study

Process overview:

• Prompt Gemini with examples of real vs. fake events.
• AI generates explanations and classifications.
• Human astronomers refine uncertain cases.

This innovation reduces reliance on massive datasets, vital for UK-led transient surveys.

Jodrell Bank and the SKA: AI Tackles Exascale Data Challenges

At the University of Manchester's Jodrell Bank Centre for Astrophysics, home to SKA Observatory headquarters, AI is essential for handling the SKA's projected data rates—terabytes per second from thousands of antennas in Australia and South Africa. SKA will map the radio sky with unprecedented detail, probing first stars, galaxies, and fundamental physics.

Manchester researchers develop machine learning for calibration, imaging, and anomaly detection amid interference. A December 2025 project led by a Manchester astronomer aims to create the most accurate radio sky model ever, using AI to process vast datasets. Recent workshops, like the February 2026 AI in Radio Astronomy event at Royal Observatory Edinburgh co-organized with Jodrell Bank, underscore UK leadership.

Key AI applications in SKA:

• Real-time beamforming and calibration.
• Fast radio burst detection.
• Predictive pulsar timing.

For aspiring researchers, opportunities abound in higher ed research jobs focusing on AI astrophysics.

Photo by Solomon Yu on Unsplash

International Collaborations: UK-South Africa Intelligent Observatory

The STFC Hartree Centre's partnership with South African Astronomical Observatory (SAAO), funded by UKRI, deploys AI for telescope self-monitoring, fault prediction, and data correction. Tools summarize observations, flag glitches, and use LLMs for document search, hosted at Hartree's supercomputing facilities.

Initiated by former astronomers Dr. Adriano Agnello and Dr. Rob Firth, it targets SALT integration next. This embeds AI in operations, broadening access and training African talent—mirroring UK university efforts in translational AI.UKRI Intelligent Observatory details

Emerging Innovations Across UK Institutions

Cambridge's Institute of Astronomy uses AI on Gaia data to spot planet-engulfing stars. University of Birmingham secured £610k with Alan Turing Institute for AI space weather prediction. Heriot-Watt University's AI reconstructs 3D black hole movies from sparse data, published December 2025. These projects highlight interdisciplinary AI-astronomy teams in UK higher education.

Benefits include:

• 85-99% noise reduction in alerts.
• Real-time event capture.
• Scalable to exascale computing.

Challenges: Data Volume, Ethics, and Compute Needs

AI demands immense compute; SKA alone requires exascale systems. Ethical issues like bias in training data and explainability persist—Oxford's astronomer panels address this. UK unis invest in trustworthy AI, balancing speed with accuracy.

Future Outlook: LSST, ELT, and Beyond

VRA adapts for LSST's 500 petabytes, while SKA early science starts 2027. European Extremely Large Telescope (ELT) will leverage UK AI for adaptive optics. UKRI funding ensures leadership, fostering PhDs and postdocs.

Photo by Conner Baker on Unsplash

Boosting Careers in AI Astronomy

UK universities offer roles blending astrophysics and machine learning. Explore higher ed jobs, lecturer jobs, or professor jobs in this field. Craft your academic CV for success. Rate professors via Rate My Professor.

Conclusion: UK Pioneering the Intelligent Universe

AI for smarter telescopes positions UK universities as global leaders, unlocking cosmic secrets efficiently. From Oxford's classifiers to Manchester's SKA pipeline, these advances promise transformative discoveries. Stay updated and pursue opportunities at university jobs, higher ed jobs, and career advice.