Tannistha Maiti is co-founder and Head of Engineering at DeepKapha AI Labs, where she leads EarthScan's applied geoscience research and the engineering behind its deep-learning systems for the upstream energy industry. She holds a PhD in Geophysics from the University of Calgary. Her work focuses on the long-tail data realities of subsurface analytics: noisy raster well logs, multi-vintage seismic acquisitions, and the practical challenges of getting models into production.
She has published extensively on:
- Reservoir characterisation: Bayesian / Hamiltonian Monte Carlo approaches to the inverse problem, integrating well logs with seismic to recover subsurface properties.
- Well-log automation: VeerNet AI, the convolutional architecture underpinning ES Raster Digitizer, plus traditional + borehole-image-log correlation pipelines.
- Seismic facies analysis: domain-adaptation methods that close the gap between annotated and unannotated 3D volumes.
- Bio-inspired materials & lignin AI: applying multimodal AI to the search for sustainable polymers from biomass.
Tannistha's research drops on EarthScan are written for practitioners: geoscientists, petrophysicists, and ML engineers who actually have to ship the work. She has a particular distaste for benchmark-only papers that don't survive contact with real field data.