This paper presents JalSwarm, an autonomous robotic system designed for effective and scalable collection of floating waste in aquatic environments. Addressing limitations in conventional waste management methods, JalSwarm utilized artificial intelligence (AI)-powered swarm robotics to ensure efficient task allocation, coordinated navigation, and dynamic obstacle avoidance. Each robot, constructed from eco-friendly materials including recycled plastic bottles and plywood, leveraged a propulsion system controlled by ESP8266 NodeMCU microcontrollers and brushless DC motors, ensuring stability, buoyancy, and cost-effectiveness. Central to JalSwarm's efficiency was its decentralized swarm intelligence algorithm, integrating YOLO-based object detection, Greedy strategies, Artificial Potential Fields (APF), and dynamic recalculation methods. This allowed robots to autonomously prioritize high-waste regions, optimize paths, and manage collision risks effectively. Additionally, a centralized monitoring interface facilitated real-time supervision, enhancing operational reliability and ease of maintenance. Experimental evaluations demonstrated a strong balance between computational efficiency and practical performance, achieving high accuracy in waste retrieval and moderate collision rates. The project's practical methodology and sustainable design highlighted significant potential for large-scale, real-time deployments in diverse aquatic settings. Ultimately, JalSwarm not only offered an innovative solution to aquatic pollution but also established a foundational framework for future advancements in autonomous environmental conservation.