Deepak Kumar

I work at the intersection of materials science and machine learning, building models and tools that make physical sense of microstructural data, process–property relationships, and extreme climate events.

I started my research journey at IIT Madras, graduating with an MS by Research in Materials Science. My thesis investigated friction stir welding (FSW) of dissimilar materials — specifically AA6082-T6 aluminium alloy and DP780 dual-phase steel — and their mechanical and microstructural characterisation using SEM, EBSD, XRD, and nano-indentation.

In parallel, I collaborated with the Oxford Nano Analysis Group (University of Oxford) to automate nano-porosity detection in TEM images using computer vision. Together we developed and published PoroDet, an open-source Python package for segmenting nano-porosities in Fresnel-contrast TEM images, designed to train on limited datasets using a U-Net CNN architecture.

Currently, I am working as a Data Scientist at IIT Delhi on high-dimensional spatio-temporal ocean data, detecting and forecasting extreme marine heatwave events using physics-aware ML/DL approaches.

Research interests: AI/ML for materials characterisation · computer vision for microstructural analysis · physics-informed modelling · materials property prediction · AI-accelerated materials discovery for energy applications

I am actively looking for research opportunities in computational materials science — particularly groups working on AI for materials discovery, microstructure–property modelling, or computer vision applied to materials characterisation.