.Engineered an automated audio processing pipeline leveraging Azure OpenAI, and Cognitive Services for transcription, diarization, translation, and PII redaction.

.Built ML-driven sentiment analysis and risk detection systems with high accuracy for compliance monitoring.Designed and implemented a RAG-based NLP system with semantic retrieval and hybrid query execution for conversational analytics.

.Designed and implemented a RAG-based NLP system with semantic retrieval and hybrid query execution for conversational analytics

Building on prior work in RF-based neural network watermarking, my project advances the design of secure and robust deep learning models. I integrate CNN, LSTM and Transformer hybrids to model multi-modal RF and sensor data, embed watermarking techniques for ownership verification, and apply optimisation strategies to preserve watermark integrity under adversarial attacks and model compression. The work also explores federated learning with client-specific watermark injection and verification.

Estimating object hardness using the gel sight sensors in robotics using raspberry and 3D image processing to obtain the high resolution of the tactile images.

Built and tested experimental frameworks simulating real-world conditions, applying machine learning models to

evaluate system robustness, achieving 85% improvement in fault response accuracy

• Experimental frameworks involved where rigorous testing procedures were executed to simulate the real world

scenarios and assess the system’s robustness.

• Applied classification and anomaly detection techniques to identify faults in real time, reducing latency by 35% and improving system reliability by over 25%.