Pyxidis builds high-performance pipelines for astronomical survey data — exoplanet detection, stellar classification, and transient event identification. Processing terabytes of sky data with NVIDIA GPUs and AI.
From raw telescope images to classified exoplanets — Pyxidis handles the full pipeline with AI-accelerated processing on NVIDIA hardware.
Transit photometry and radial velocity pipelines with deep learning models. Detect exoplanet candidates from light curves in seconds. CNN + transformer models trained on TESS and Kepler data. 99.4% precision.
Automated spectral classification from O to M type. CNN models trained on 4M+ spectra from SDSS, LAMOST, and Gaia. Classify 1M stars per hour with spectroscopic and photometric fusion.
Real-time transient event detection in survey streams. Supernovae, kilonovae, and AGN flares identified within minutes of observation. GPU image subtraction and ML classification in one pipeline.
Automated galaxy classification and morphological analysis. CNN models identify spiral, elliptical, and irregular galaxies from survey images. Star formation rate estimation from spectral energy distributions.
End-to-end processing: calibration, stacking, source extraction, photometry, and catalog generation. Handles LSST-scale data rates (20TB/night) with distributed GPU processing across clusters.
ML-powered star and galaxy catalogs with uncertainty quantification. Cross-match across surveys (Gaia, 2MASS, WISE). Automated quality flags and anomaly detection for data integrity.
From telescope raw frames to classified catalogs — a multi-stage pipeline optimized with NVIDIA GPUs.
Raw telescope frames ingested via FITS. GPU-accelerated bias/dark/flat correction, astrometric calibration, and cosmic ray removal. Processes 20TB/night at LSST scale.
GPU-accelerated source detection, aperture photometry, and PSF fitting. Source Extractor on GPU. ML-based star/galaxy separation. 50M sources per hour per GPU.
Deep learning models for exoplanet, stellar, and galaxy classification. Transformer models for light curve analysis. Outputs to searchable catalog with cross-matching.
Deployed at observatories, research institutions, and space agencies.
Processing TESS full-frame images in real-time. 12,000+ exoplanet candidates detected, 480 confirmed. Transit detection in <30 seconds per light curve. Used by 15 research institutions.
Prototype pipeline for Vera C. Rubin Observatory. Handles 20TB/night data rate with distributed GPU processing. Source extraction and calibration 40x faster than CPU baseline.
Real-time transient detection on Zwicky Transient Facility streams. Supernovae classified within 4 minutes of first light. 3,200 transients detected and classified in 2025.
Cross-matching 1.7B Gaia sources with ground-based surveys. ML-powered stellar parameter estimation (Teff, log g, [Fe/H]) for 200M stars. Catalog published to SIMBAD.
Join observatories and research teams using Pyxidis to accelerate astronomical data processing, exoplanet detection, and sky survey analysis.