Coal India Demos

Semantic retrieval and grounded generation over Detailed Project Report excerpts, with structured data extraction. Answers cite sources inline and decline rather than guess.

Automated per-pixel change assessment of mine reclamation from satellite imagery, 2016 to 2025. Hotspot ranking and unsupervised clustering surface what to investigate.

A trained, calibrated ensemble model predicting coal quality with verified 95% confidence bands. Batch scoring flags exceptions and detects lab bias.

An AI judge checks each environmental clearance condition against evidence with citations and a strict rubric. Also drafts compliance memos and evidence-request letters.

Neural search over incident narratives finds matches keyword search misses. An AI analyst summarizes patterns and drafts fleet safety briefings.

Grounded, cited answers over coal block auction tranche documents, with computed ratios shown transparently. Generates one-click tranche briefing notes.

Every dispatch scored against its own block's statistical baseline the moment it's logged. Sigma-flagged exceptions get an AI-drafted resampling-request memo, never an automatic call.

Retrieves and cites what a mine's Mining Plan actually sanctions, computes the real OB:coal trend with the arithmetic shown, and only drafts a variance note when the fitted trend genuinely warrants one.

The Kuznetsov fragmentation formula feeds a trained model that corrects it against real shovel-productivity outcomes. Flags underperforming faces and drafts the shift deployment call.

Itemizes every cost (compute, cloud vs on-prem, talent, integration) against quantified benefits, then computes NPV, IRR, and payback across optimistic/base/pessimistic scenarios - a cost only looks good next to a credible benefit.