Research Projects

🏆 $750k Award Winner - Developed an AI-driven intersection safety system that secured Tier 1 funding from the U.S. Department of Transportation.

Key Achievements:

• Designed and implemented a multi-modal perception system for intersection monitoring
• Developed real-time collision prediction algorithms with 95%+ accuracy
• Created a comprehensive testing framework using CARLA simulation environment
• Led a team of 5 researchers and coordinated with industry partners
• Published preliminary results and filed 2 provisional patents

Technical Impact: The system can predict potential collisions 3-5 seconds before they occur, providing critical time for intervention. Our approach combines LiDAR, camera data, and V2X communications to create a comprehensive safety net for urban intersections.

🏆 Best Paper Award at IEEE ITSC 2024 - Developed novel algorithms for understanding and predicting vehicle movement patterns in urban environments.

Key Contributions:

• Designed a graph-based representation for semantic trajectory understanding
• Implemented attention mechanisms for multi-scale temporal pattern recognition
• Achieved 15% improvement over state-of-the-art trajectory prediction methods
• Processed and analyzed over 10TB of real-world GPS trajectory data
• Collaborated with city planning departments for real-world validation

Real-world Impact: Our methods are being piloted by 3 major cities for traffic optimization and urban planning. The semantic understanding enables better infrastructure decisions and more efficient traffic management.

🎯 High-Impact Solution - Implemented YOLOv5-based mask detection system to ensure COVID-19 safety compliance across university campus and public spaces.

Technical Achievements:

• Achieved 97% accuracy in real-time mask detection across diverse demographics
• Optimized model for edge deployment with 30fps performance on Jetson Nano
• Integrated with existing security camera infrastructure
• Deployed system across 15+ campus locations during peak COVID period
• Developed privacy-preserving detection that doesn't store facial data

Social Impact: The system was instrumental in maintaining campus safety during the pandemic, processing over 100,000 detections daily and helping enforce health guidelines while respecting privacy.

🧠 AI for Interactive Storytelling - Collaborated on developing constrained text generation methods for creating coherent narratives in text adventure games.

Research Contributions:

• Designed knowledge graph integration methods for transformer-based language models
• Improved narrative consistency by 40% compared to unconstrained generation
• Contributed to novel evaluation metrics for story coherence
• Implemented efficient graph attention mechanisms for real-time generation
• Presented findings at NAACL 2022 Workshop on Wordplay

Innovation: Our approach enables AI systems to generate more coherent and contextually appropriate text by leveraging structured knowledge, opening new possibilities for interactive entertainment and educational applications.

Developing the next generation of autonomous driving systems that leverage large language models for better reasoning and decision-making in complex traffic scenarios.

Expected Impact: This project aims to create more interpretable and adaptable autonomous vehicles that can handle edge cases through natural language reasoning.