MegaMiner flagship

MegaMiner Shadow Lab

Every weird light curve is a shadow asking for a geometry. Switch between ordinary planets, Clarke exobelts, dust, disintegrating bodies, structured occultations, and L1 starshades while the matching curve updates in real time.

Shadow Lab Bezier Simulator Analysis progress

Automated search graph

MegaMiner pipeline from TESS threshold-crossing events to vetted anomaly candidates.

The operational pipeline combines ExoMiner scoring, shape-anomaly statistics, a conditional autoencoder for eclipsing-binary rejection, and LLM-assisted literature and DV-report vetting.

MegaMiner on TARANG Compute node for large-scale folded TESS light-curve analysis

Agentic-AI orchestration Coordinates ExoMiner scoring, anomaly filtering, literature queries, and DV-report review

Original sample TESS TCEs

Approximately 1 million threshold-crossing events from the public TESS archive.

Manual cuts Quality screen

Quality masks, sector stitching, catalogue matching, and basic sky-localization checks.

ExoMiner + Z-score Odd-shape score

Planet, astrophysical false-positive, instrumental-artifact, and transit-shape anomaly probabilities.

Conditional autoencoder Binary rejection

Purpose-built anomaly model removes eclipsing binaries and recurring astrophysical contaminants.

LLM-based triage Context vetting

Literature, catalogues, and TESS DV reports are queried before candidates move to human review.

Search publications + catalogues

Generate candidate packets with archive provenance, known-object matches, and review-ready evidence.

Anomaly vocabulary

Features MegaMiner would score before deciding what they mean.

Our MegaMiner pipeline is framed to be skeptical by design. Every strange signal is first a candidate morphology, then an astrophysical or instrumental investigation, and only much later a technosignature discussion.

Ingress and egress Symmetry, sharpness, tails

Planets tend to be smooth and repeatable; dust and structures can leave asymmetric edges.

Depth evolution Stable, changing, or fragmented

Multiple sectors can reveal whether a dip is coherent, precessing, evaporating, or transient.

Recurrence logic Periodic, quasi-periodic, or single event

The cadence of shadows can separate orbital systems from stellar activity and artifacts.

Geometric texture Smooth curves versus hard structure

Step-like or multi-component dimming motivates deeper checks against known populations.

Candidate discipline

A strange dip warrants scrutiny, not a verdict.

MegaMiner's strongest scientific posture is not to declare artificiality. It is to make anomalies reproducible, comparable, and hard to ignore when standard categories fail.

Archive provenance

Mission, sector, aperture, quality flags, crowding, and nearby contaminant checks.

Known populations

Planet transits, eclipsing binaries, pulsators, spots, flares, exobelts, dust, and debris.

Instrumental vetoes

Momentum dumps, scattered light, detector artifacts, background events, and pipeline edges.

Follow-up packet

Folded and raw light curves, target metadata, comparable analogs, and next-observation value.