Real-Time Plankton Monitoring with RAPID

As a researcher at the Alan Turing Institute, I contributed to the development of RAPID, a suite of real-time analytics tools designed for the Plankton Analytics Pi-10 instrument. Deployed at sea, this tool suite enables in situ monitoring and classification of marine plankton through machine learning models, allowing for rapid data-driven insights on ocean biodiversity.

Project Highlights

• Real-Time Analysis: Developed and optimized models for immediate plankton image classification, integrated with Edge-AI to facilitate on-site processing.
• Cross-Institution Collaboration: Worked alongside the National Oceanography Centre to design sensor-agnostic models, enabling stratification of marine particles across multiple sensor types.
• Impact on Marine Research: The tool enhances our understanding of marine ecosystems, supporting conservation efforts by improving long-tail classification in plankton data.

Technical Details

• Machine Learning Methods: Leveraged advanced techniques such as convolutional neural networks (CNNs) and metadata-enhanced representation learning to achieve high accuracy in image classification.
• Tools & Frameworks: Utilized TensorFlow, PyTorch, and Scikit-learn for model development, with deployment supported by a custom API that provides real-time access to processed data via a live dashboard.

This project exemplifies my expertise in environmental AI and underscores my commitment to applying machine learning in biodiversity monitoring.