Portfolio item number 1
Short description of portfolio item number 1
Short description of portfolio item number 1
Short description of portfolio item number 2 
Working Paper, 2026
Introducing a biological receptor-glomerular bottleneck modeling olfactory priors into sequence models, boosting low-resource NER performance by +6.23% F1 in Bangla and +4.31% F1 in Tamil.
Recommended citation: Deshpande, B. (2026). Olfactory-Inspired Sparse Combinatorial Coding for Low-Resource Named Entity Recognition. arXiv preprint arXiv:2606.21895.
Download Paper
Under review, ICLR 2027, 2026
A dynamic structural sparsification algorithm based on biological synapse elimination, achieving 72.79% structural sparsity on VGG-16 while improving CIFAR-10 baseline accuracy by +1.70%.
Recommended citation: Deshpande, B. (2026). Dynamic Activity-Dependent Pruning (DADP): A Reverse-Hebbian Structural Plasticity Framework for Neural Networks. arXiv preprint arXiv:2610.11853 (Under review, ICLR 2027).
Download Paper
Published:
Invited research presentation discussing data curation, synthetic pipelines, and high-EQ modeling distributions for low-resource Indic languages.
Published:
Invited talk breaking down custom curriculum training solutions for structural LID tasks in code-mixed and transliterated language environments.
Undergraduate course, University 1, Department, 2014
This is a description of a teaching experience. You can use markdown like any other post.
Workshop, University 1, Department, 2015
This is a description of a teaching experience. You can use markdown like any other post.