Projects

Adversarial Defense in Collaborative Vehicular Perception

January 2025 - March 2025
Python Scikit-learn PyTorch LLM
  • Developed an uncertainty-aware adversarial defense system for collaborative vehicular perception using PyTorch, implementing and diffusion models, achieving 88% detection accuracy with 30-35ms latency
  • Integrated a robust consensus mechanism (ROBOSAC-inspired) with adaptive thresholding to identify adversarial attacks across multiple vehicles, maintaining a low false positive rate of 7.7-10.4%

Meal Nutrition Analysis using Multi-Modal Data

October 2024 - December 2024
Python Scikit-learn PyTorch
  • Developed a multimodal deep learning model to accurately estimate calorie intake using a dataset of 25K samples, achieving a 30% improvement in prediction accuracy over traditional methods
  • Integrated CNNs, bidirectional LSTMs with attention mechanisms, and fully connected networks. Achieved a Root Mean Square Relative Error (RMSRE) of 0.35