Real-Time Traffic Detection System using YOLO

Developed a real-time traffic monitoring system using YOLO for object detection on live video streams. Implemented detection of vehicles and traffic elements with optimized inference pipelines for real-time performance. Designed a system capable of processing continuous video input, detecting and tracking objects with low latency. Applied model optimization techniques to improve detection speed while maintaining accuracy in dynamic environments. Handled full pipeline including data preprocessing, model integration, and real-time visualization of detection results. Technologies: Python, YOLO (v5/v8), OpenCV, PyTorch, real-time video processing.

Multimodal AI Chatbot with RAG and Document Understanding

Developed a multimodal AI chatbot supporting text, documents, and image inputs using Retrieval-Augmented Generation (RAG). Implemented a pipeline combining LLMs with document retrieval to generate context-aware responses and reduce hallucination. Designed scalable data processing workflows for handling structured and unstructured inputs. Built interactive interface for real-time querying and response generation. Technologies: Python, LLMs, RAG, NLP, Streamlit.

AI-Powered Teleoperation System for Humanoid Robot (Pepper)

Designed and developed a real-time AI-powered teleoperation system for a humanoid robot (Pepper) using computer vision and multimodal perception. Built end-to-end pipelines integrating object detection, face recognition, gesture recognition, depth estimation, and speech processing into a unified system. Engineered low-latency communication and perception pipelines achieving real-time performance (<50ms), enabling smooth remote interaction using VR-based control interfaces. Handled full system integration including AI inference, control logic, and real-world deployment, ensuring stability under continuous operation. Technologies: Python, PyTorch, OpenCV, YOLO, MediaPipe, real-time systems integration.