Dr Benjamin O'Driscoll

Ben joined PML in April 2022 shortly after completing his PhD in Graphene Biosensing from the University of Plymouth. Throughout his studies, Ben developed a passion for the automated processing, analysis and visualisation of data, skills that he has continued to develop in his current role at PML within the Environmental Intelligence group.

Ben is a trained physicist, studying Physics (BSc) at the University of Warwick between 2010-2013. After his undergraduate degree, Ben worked in industry as a marine sensing analyst where he embraced all things data. In his current role, Ben enjoys the challenge of handling vast collections of data, producing useful data insights and presenting these in a user-friendly manner.

Key Projects

• 2025-Present – Phytodiverse (ESA)
  • Developed a full machine learning workflow building a Random Forest Regressor model to predict phytoplankton pigments from hyperspectral measurements
  • Prepared and preprocessed hyperspectral datasets by cleaning data, handling missing values and creating training, validation and test datasets.
  • Benchmarked the Random Forest model against alternative approaches to assess predictive performance and computational efficiency
  • Performed systematic hyperparameter tuning using cross-validation to optimise model parameters
• 2025-Present – ISOLUME – JPI Oceans (DEFRA)
  • Built scalable pipelines to process historic remote sensing datasets (OCCI), analysing temporal changes in inherent and apparent optical water properties.
  • Leveraged high-performance computing clusters to efficiently handle large, multi-year datasets, applying statistical models to detect trends and variability over time.
  • Developed visualisations and data products to communicate changes in water properties, supporting interpretation and decision-making for environmental monitoring.
• 2025-Present – DeepVision 
  • Built end-to-end computer vision machine learning pipeline for benthic marine imagery, automating data ingestion via API, preprocessing, and object detection workflows
  • Applied containerised YOLOv8 object detection models to identify and classify seabed features, optimising performance for large-scale image datasets
  • Streamlined the annotation lifecycle by programmatically uploading generated labels back to the platform, improving efficiency and reducing manual effort
• 2025-Present – Assisted Planning Addressing Climate Effects (ASPACE) – NERC / INSITE/ Horizon Europe / AXA Investment Managers
  • Developed the AI-driven component for the platform, optimising area selection based on environmental and user-defined constraints
  • Implemented economic models to evaluate planning scenarios, integrating user-supplied data to generate adaptive, data-informed recommendations
  • Integrate the AI pipeline into an interactive platform, enabling users to explore, assess, and refine spatial planning outcomes
• 2024-2025: Aquatic Pollution from Light and Anthropogenic Noise (AquaPLAN) (EU)
  • Leveraging xarray and matplotlib packages to efficiently process, manipulate and visualise large geospatial datasets.
  • Scalable scripts developed to optimise processing across compute clusters using SLURM to manage resources.
• 2023-2025: ProBleu Catalogue of Teaching Aids (EU)
  • Fullstack application involving file translation pipeline used to host multi-lingual water literacy materials.
  • Application harnesses Django to manage back-end server-side logic, communicating with a PostgreSQL database whilst Celery provides asynchronous processing support.
• 2024-2025 – Synchronising Earth Observation and Modelling Towards a Digital Twin Ocean (SyncED-Ocean)
  • Digitial twin demonstrator combining data from satellite Earth Observations, marine autonomous platforms and marine system models
  • Operational processing involving the continual downloading of asset datasets
  • Post mission development of visualisation portal for high level decision support tool.
• 2024: eDNA QC Application (Data Map)
  • Data science R and R Shiny applet facilitating the manipulation of eDNA datasets to conduct data exploration, analysis and visualisation.