AVICA : A fully automated CASA pipeline for large-volume very long baseline interferometry data calibration

Calibrating large volumes of very long baseline interferometry (VLBI) data is a time-consuming process, traditionally requiring significant human intervention at every stage, from data inspection and parameter tuning to calibrator selection. While the Common Astronomy Software Applications ( ) package is the standard data reduction tool at major radio observatories, no existing -based pipeline is capable of operating in a fully automated manner across the heterogeneous data formats produced by the Very Long Baseline Array (VLBA) over three decades of operations. The Search for Milli-Lenses (SMILE) project requires the calibration of sim5000 VLBA sources, which makes such blind automation a practical necessity. CASA CASA We introduce the Automated VLBI pipeline in ( ). It is designed to automate the calibration of VLBA datasets for the SMILE sample, and its functionalities are available for the broader VLBI community. CASA AVICA AVICA extends the existing -based calibration framework by automating the preprocessing of archival VLBA data, the selection of calibrators and reference antennas, and the execution of the full calibration workflow. Preprocessing operates on both the FITS-IDI and measurement set data formats, extracting only the required sources from large archive files to reduce data volume and loading time. Calibrators and reference antennas are ranked automatically using the signal-to-noise ratio from the fast Fourier transform-based fringe detection, and the resulting parameters are passed to for calibration. Workflow management and progress tracking are handled by Automated Logical Framework for executing Dynamic scripts ( ), which orchestrates pipeline execution for each dataset and records results to an external spreadsheet in real time. CASA rPICARD rPICARD ALFRD AVICA was validated on a sample of 1000 sources with NRAO archival data spanning observations from 1995 to 2023, for a total of 1372 individual band-separated observations across the S, C, X, U, and K bands. The pipeline successfully produced calibrated output for 978 sources (97.8%), with the 22 failures attributable to corrupted or incomplete input data. The mean per-source execution time across the 1000-source sample was approximately 30 minutes when using Message Passing Interface parallelization with up to 20 cores. AVICA demonstrates that a fully blind calibration of heterogeneous archival VLBA data is achievable using without manual parameter input. Although validated on archival VLBA data, the underlying algorithms are designed to be generalizable to other VLBI arrays, and the automated calibrator and reference antenna selection will be incorporated into a future release, extending blind-automated calibration to any supported array. and are available as open-source Python packages. CASA rPICARD AVICA ALFRD