This project develops a diffusion-based synthetic data and defect generation platform to enhance AI-driven manufacturing. It uses advanced generative models to create high-quality, labeled images that replicate real-world variations and rare defect scenarios, addressing the challenge of limited or sensitive production data. By enabling scalable dataset generation, the platform significantly accelerates training of computer vision models used in quality inspection and predictive maintenance. It integrates seamlessly with existing industrial pipelines, allowing organizations to adopt the solution without major infrastructure changes. The approach reduces data acquisition costs while improving model robustness and accuracy across diverse conditions. Overall, the platform supports faster deployment of AI systems, improves operational efficiency, and strengthens decision-making in manufacturing environments by providing reliable, diverse training data at scale.

The objective of this research is to predict larval habitat threat maps and identify potential future mosquito breeding grounds, particularly areas likely to retain water. A portion of the dataset has been carefully annotated to support model development. Benchmark studies and deployment of off-the-shelf models have shown promising initial results. Further data annotation and model training are currently in progress. The broader plan is to pursue joint technology development, enabling opportunities for research grants as well as new industry collaborations.

Funded through NM-ICPS, this project aims at developing a unified AI platform combining multi-omics integration, graph transformer-based network inference and LLM-driven knowledge analytics to prioritize Alzheimer’s disease genes. The technology identifies key regulatory and peripheral genes with high biological relevance, offering interpretable outputs for precision medicine, biomarker discovery, and accelerated therapeutic target exploration.

Smart Prediction with Infrared Detection of Residual cancer (SPiDR) is a fully functional proof‑of‑concept intraoperative device that uses active thermal imaging and an AI‑driven image analysis pipeline to detect residual cancer tissue in real time during surgery. This has been already validated at Sagore Dutta Hospital, Kolkata