HUB2026

Spectral, spatial, and hyperspectral data—whether obtained from astronomical observatories or from terrestrial and orbital remote-sensing platforms—encode intricate mixtures of physical signals, all shaped by noise, calibration uncertainties, and instrument-specific distortions. Across these domains, researchers rely on a wide range of analytical tools: from classical statistical inference and signal processing to modern machine-learning, geometric, and physics-informed methods.

Despite the diversity of applications: e.g. galaxy evolution, exoplanet atmospheres, environmental monitoring, precision agriculture, mineral mapping. The underlying challenges are remarkably similar, high-dimensional measurements, correlated noise, nonlinear mixing of sources, and the need for robust spatial–spectral modeling.

This common structure creates a natural bridge between astronomy, Earth observation, and applied data science. Techniques developed in one field (e.g., spectral unmixing, denoising, segmentation, dimensionality reduction, latent-space modeling, uncertainty quantification) often transfer seamlessly to the others.

Framing the problem space in this unified way provides a strong rationale for a cross-disciplinary hack week: a setting where participants from different backgrounds can exchange methods, compare workflows, and co-develop tools that generalize across spatial and hyperspectral datasets—regardless of whether the telescope is pointed outward toward galaxies or downward toward Earth.

This hack week, held on the beaches of Itacaré in Bahia (Brazil), aims to bring together experts from diverse fields to prototype computational solutions inspired by—but not limited to—an astrophysical challenge: the extraction of point-source spectra embedded in noisy hyperspectral data, a problem central to Integral Field Spectroscopy.

One of the flagship projects will focus on disentangling the light of some of the oldest stellar systems known to humankind: the globular clusters. By tackling this problem in a cross-disciplinary setting, participants will explore techniques that generalize well beyond astronomy—into remote sensing, environmental monitoring, precision agriculture, materials science, and any domain working with spatial–spectral data affected by complex noise and source mixing.

Datasets from other fields are welcome! While we are open to any type of expertise participants may bring, we list below some examples of research topics we consider particularly synergistic with the goals of the hack week: