• QUBO formulation of classical learning models
• Quantum annealing-based optimization
• Applications for 3D Computer Vision tasks

• Object recognition pipeline with 80% accuracy for optimized inventory management
• Unsupervised learning technique for real-time object localization and segmentation from retail shelf images
• Embeddings from a pre-trained model to analyze and identify similar objects in the unlabeled dataset

• Custom dataset curation for ocular biomarkers with clinician-verified ground truth
• ML-based diagnosis of Diabetic Retinopathy and Hypertensive Retinopathy with accuracy rate of 98%
• SHAP-enabled explanation of biomarker impact on the diagnosis prediction

• Course: AI for Everyone (20IC206T)

Screening Systemic Disorders via Retinal Vascular Biomarkers

• Clinician-in-the-loop deep learning and image processing pipeline for computing retinal parameters
• Results within ±8% of the research benchmark SIVA
• Decision support system plan to address the confounding factors in grading Retinopathy of Prematurity
• Review of clinical AI-based retinal imaging telecare services in India to improve care outreach

Context-Sensitive Semantic Search Tool

• Semantic search tool for impact analysis in software testing with 95% accuracy
• Embeddings using a pre-trained language model (ELMo) for semantic mapping of test cases
• Interactive UI visualization scheme for the retrieved results