Currently researching applications of deep reinforcement learning in cybersecurity of cyberphysical systems.

Developed Smart Warehouse Surveillance Suite for the UN World Food Programme, enhancing real-time surveillance across 10 warehouses in India.1) Established a hardware agnostic docker-based video analytics pipeline for scalable, multi-class AI inference, processing multiple CCTV feeds with robust stream multiplexing and batching.2) Trained and Optimized models for Jetson and discrete GPUs, balancing efficiency and latency, while analyzing trade-offs between central and edge processing paradigms.3) Integrated media servers and bash scripts to streamline data collection, cleaning, augmentation and remote collaboration for efficient annotation and processing of camera feeds.

Machine Intelligence & Bio-Motion (MIBM) Research Group. Host - Prof. Anup Nandy1) Developed an EEG data processing pipeline, including filtering, epoch extraction, event labeling, and spectrogram visualization using STFT.2) Conducted literature review and analyzed correlations between EEG frequency bands and memory processes such as working and episodic memory.3) Performed data transformations, applied models (CNN, LSTM, Liquid Neural Nets), and evaluated results to investigate cognitive processes during memory encoding and decoding.

Lead a small team and worked on a research project proposing a novel approach to map mineral deposits, leveraging the Geological Survey of India’s geological, geophysical, and geochemical data repositories along with the innovative use of convolutional and generative neural networks.

1. Developed and tested various ML models to predict equipment failures, optimizing reliability, reducing downtime, and enhancing maintenance strategies for critical machinery.2. Created a novel hybrid CNN-LSTM architecture with an exceptional MSE of about 15 for RUL prediction of engines.3. Implemented a system that utilizes GANs for data augmentation, enabling quick modelling and deployment even when faced with limited data resources.

1. Worked with teams on Computer Vision systems and Deep Learning architectures to extract information from CCTV feeds for improved safety and productivity at client refineries.2. Worked with CI/CD tools like Docker and Jenkins to streamline the application development workflow.