Magnetic AB stars are known to produce periodic radio pulses by the electron cyclotron maser emission (ECME) mechanism. Only 19 such stars, known as ‘Main-sequence Radio Pulse emitters’ (MRPs), are currently known. The majority of MRPs have been discovered through targeted observation campaigns that involve carefully selecting a sample of stars that are likely to produce ECME and which can be detected by a given telescope within reasonable amount of time. These selection criteria inadvertently introduce bias in the resulting sample of MRPs, which affects subsequent investigation of the relation between ECME properties and stellar magnetospheric parameters. The alternative is to use all-sky surveys. Until now, MRP candidates obtained from surveys were identified based on their high circular polarisation ($\gtrsim 30\%$). In this paper, we introduce a complementary strategy, which does not require polarisation information. Using multi-epoch data from the Australian SKA Pathfinder (ASKAP) telescope, we identify four MRP candidates based on the variability in the total intensity light curves. Follow-up observations with the Australia Telescope Compact Array (ATCA) confirm three of them to be MRPs, thereby demonstrating the effectiveness of our strategy. With the expanded sample, we find that ECME is affected by temperature and the magnetic field strength, consistent with past results. There is, however, a degeneracy regarding how the two parameters govern the ECME luminosity for magnetic A and late-B stars (effective temperature $\lesssim 16$ kK). The current sample is also inadequate to investigate the role of stellar rotation, which has been shown to play a key role in driving incoherent radio emission.

This work introduces GalProTE, a proof-of-concept Machine Learning model, leveraging Transformer Encoder architecture to efficiently determine the stellar age, metallicity, and dust attenuation of galaxies from optical spectra. Designed to address the challenges posed by the vast datasets produced by modern astronomical surveys, GalProTE offers a significant improvement in processing speed while maintaining accuracy. Using the E-MILES spectral library, we generate a dataset of 111936 diverse templates by expanding the original 636 simple stellar population models with varying extinction levels, combinations of multiple spectra, and noise modifications. This ensures robust training over the spectral range of 4750–7100 Å at a resolution of 2.5 Å. GalProTE architecture employs four parallel attention-based encoders with varying kernel sizes to capture diverse spectral features. The model demonstrates a mean squared error (MSE) of 0.27% with a standard deviation of 0.10% between the input spectra and the GalProTE-generated spectra for the synthetic test dataset. Performance evaluation against real data from two galaxies in the PHANGS-MUSE survey (NGC4254 and NGC5068) demonstrates its ability to extract physical parameters efficiently, with spectral fit residuals showing a mean of -0.02% and 0.28%, and standard deviations of 4.3% and 5.3%, respectively. To contextualize these results, we compare age, metallicity and dust attenuation maps generated by GalProTE with those of pPXF, a state-of-the-art spectral fitting tool. While pPXF achieves robust results, it requires approximately 11 sec per spectrum. In contrast, GalProTE processes a spectrum in less than 4 ms – a speedup factor exceeding 2750, while also consuming 68 times less power per spectrum. The comparison with pPXF maps from PHANGS-MUSE underscores GalProTE’s capacity to enhance traditional methods through machine learning, paving the way for faster, more energy-efficient, and more comprehensive analyses of galactic properties. This study demonstrates the potential of GalProTE as an efficient, scalable, and sustainable solution for processing large astronomical surveys.

We present Evolutionary Map of the Universe Search Engine (EMUSE), a tool designed for searching specific radio sources within the extensive datasets of the Evolutionary Map of the Universe (EMU) survey, with potential applications to other Big Data challenges in astronomy. Built on a multimodal approach to radio source classification and retrieval, EMUSE fine-tunes the OpenCLIP model on curated radio galaxy datasets. Leveraging the power of foundation models, our work integrates visual and textual embeddings to enable efficient and flexible searches within large radio astronomical datasets. We fine-tune OpenCLIP using a dataset of 2 900 radio galaxies, encompassing various morphological classes, including FR-I, FR-II, FR-x, R-type, and other rare and peculiar sources. The model is optimised using adapter-based fine-tuning, ensuring computational efficiency while capturing the unique characteristics of radio sources. The fine-tuned model is then deployed in the EMUSE, allowing for seamless image and text-based queries over the EMU survey dataset. Our results demonstrate the model’s effectiveness in retrieving and classifying radio sources, particularly in recognising distinct morphological features. However, challenges remain in identifying rare or previously unseen radio sources, highlighting the need for expanded datasets and continuous refinement. This study showcases the potential of multimodal machine learning in radio astronomy, paving the way for more scalable and accurate search tools in the field. The search engine is accessible at https://askap-emuse.streamlit.app/ and can be used locally by cloning the repository at https://github.com/Nikhel1/EMUSE.