Debabrata Pruseth

Transfer learning has transformed modern computer vision by enabling powerful AI models to achieve high performance with limited labelled data and reduced training costs. This research article explores practical transfer learning techniques across CNNs, Vision Transformers, CLIP-style vision-language models, and foundation segmentation models. Topics include feature extraction, fine-tuning strategies, self-supervised learning, LoRA adaptation, prompt-based transfer, and real-world applications in healthcare, agriculture, satellite imagery, retail, and industrial inspection.

Research note on using deep learning for image-based plant disease detection, crop diagnosis, and practical AI-assisted agriculture workflows.

This research note presents a practical deep learning framework for automated skin lesion classification, showing how AI can support skin cancer detection workflows through image-based analysis.

This research note presents an end-to-end data quality management framework for banking systems, showing how generative AI can support profiling, validation, anomaly detection, root-cause analysis, and governance.

This research note explores how AI can support protein analysis, from sequence representation and structural interpretation to drug-binding hypothesis generation for early-stage drug discovery.

This research note explores how an AI-powered persona-driven survey engine can simulate user responses, model synthetic personas, and support structured decision-making through feedback analysis.

This research note explores how reasoning in large language models is evolving from linear chain-of-thought prompting to graph-based Diagram-of-Thought frameworks that support more structured, multi-path reasoning.

Can artificial intelligence learn the behaviour of fluid dynamics from simulation data? This research note explores how AI can support prediction of Rayleigh-Bénard convection using WALRUS and scientific machine learning.