Worked on a DST (Govt. of India)-funded project to build an end-to-end pipeline for Indic-language radiology report generation using a two-stage multimodal model framework. Also developed rank-based conformal prediction methods to improve reliability in few-shot pathological analysis pipelines.

Worked on a GNN-based accident detection system for a smart traffic solution for the Govt. of Odisha, cutting emergency response times by over 60%. Also designed an EHR application handling 100,000+ daily records and improved record retrieval by 42%.

Developed conformal prediction methods to enhance uncertainty quantification in pathological cell classification workflows, improving model interpretability and robustness. Also designed a sampling-based feature bias mitigation technique to address data-driven biases in cervical cytology classification, improving model fairness and reliability.

Finetuned a Llama2-13b model for a summarization platform with custom review tags for grey literature of regeneration practices on ASReview Lab.

Worked on a Contrastive Self-Supervised Learning (CSSL) approach for macromolecular structure classification from cryo-ET data with limited labels.

Participated in the 2023 Neuromatch Academy Summer School on Computational Neuroscience. Designed regression models for future motion state prediction using time series analysis on ECoG data.

Partcipated in the 3rd Edition Summer School on Deep Learning for Medical Imaging (DLMI-22). Evaluated various weakly supervised segmentation techniques for cardiac diseases diagnosis.

Worked on building the 'Indian Brain Segmentation Dataset'- a sub-cortical structure segmentation database for young population. Also contributed to a multi-scale attention architecture for COVID-19 detection from Chest-X Rays.