AI Research Intern
Scrapify Ecotech – Ecofloater Project
Led the development of a revolutionary autonomous Ecofloater system for river cleaning at Scrapify Ecotech, an environmental technology startup incubated at iTNT Hub, Anna University. Pioneered computer vision solutions using custom YOLO architectures and Vision Transformers for real-time debris detection, integrating Intel RealSense depth cameras and reinforcement learning for autonomous navigation on NVIDIA Jetson edge hardware.
2024 - 2025
6-month intensive AI research internship with hands-on system development
iTNT Hub, Anna University
Guindy, Chennai - 600015 (On-site)
Scrapify Ecotech
Environmental technology startup focused on autonomous water body cleaning solutions
CompletedAs the lead AI Research Intern at Scrapify Ecotech, I spearheaded the development of the complete autonomous intelligence stack for the Ecofloater river cleaning system. This role involved cutting-edge research in computer vision, autonomous systems, and environmental technology. I was responsible for the entire AI pipeline—from synthetic dataset generation and model architecture design to edge deployment optimization—resulting in a fully functional self-thinking cleaning system capable of real-time debris detection and autonomous navigation in dynamic aquatic environments.
Key Responsibilities
Developed custom YOLO (You Only Look Once) and Vision Transformer architectures specifically optimized for debris detection in aquatic environments. Achieved high accuracy in real-world conditions with varying lighting, water turbidity, and debris types through extensive augmentation and synthetic data generation.
Implemented sophisticated camera calibration mechanisms including stereovision and depth vision processing with Intel RealSense D435i cameras. Developed precise 3D environmental mapping algorithms for accurate debris localization and distance estimation in real-time.
Integrated reinforcement learning algorithms enabling the system to make intelligent decisions about cleaning priorities, navigation paths, and resource optimization. The system continuously learns and adapts to different environmental conditions and debris patterns without human intervention.
Successfully deployed the complete AI pipeline on NVIDIA Jetson Nano with optimized real-time inference achieving 15+ FPS. Implemented TensorRT optimization, model quantization, and memory-efficient processing for resource-constrained edge environments.
Technical Environment
Training Systems
High-end workstations with NVIDIA RTX GPUs, CUDA-optimized configurations, and 64GB+ RAM for large-scale model training and synthetic data generation.
Edge Deployment
NVIDIA Jetson Nano for primary inference, Raspberry Pi 4 for auxiliary processing, and custom PCBs for sensor integration.
Sensor Suite
Intel RealSense D435i depth cameras, custom stereo camera rigs, and environmental sensors for comprehensive situational awareness.
Deep Learning Frameworks
PyTorch for model development, Ultralytics for YOLO implementations, and Hugging Face Transformers for Vision Transformer architectures.
Computer Vision
OpenCV for image processing, Open3D for point cloud manipulation, and custom RealSense SDK integration for depth stream processing.
Optimization Tools
NVIDIA TensorRT for inference acceleration, ONNX for model portability, and custom CUDA kernels for performance-critical operations.
Technology Stack
Key Achievements
Successfully developed and deployed a fully autonomous river cleaning system with real-time debris detection achieving 95%+ accuracy in field conditions.
Achieved optimized real-time inference at 15+ FPS on NVIDIA Jetson Nano through TensorRT optimization and model quantization techniques.
Created comprehensive synthetic and real-world datasets with 10,000+ annotated images specifically designed for aquatic debris detection.
Integrated computer vision, reinforcement learning, depth sensing, and edge computing into a cohesive autonomous system architecture.
Documents & Resources
Official completion certificate from Scrapify Ecotech acknowledging AI research contributions
Original internship offer letter detailing position terms and research responsibilities
Ecofloater Autonomous System
This internship resulted in a comprehensive research project on autonomous aquatic cleaning systems. View the complete technical documentation, architecture details, and deployment specifications.
View Full Research Details