Aakash Varambhia - Delivering state of the art imaging data science to aid research and development

Learn how Johnson Matthey built a data science platform for processing terabyte-scale imaging data to optimize catalyst design, combining open source tools with user-friendly web interfaces.

Key takeaways
  • Johnson Matthey is focused on catalyst design and optimization, working at multiple scales from atomic to device level to analyze and improve catalytic materials

  • Main challenges include processing terabyte-scale imaging datasets, dealing with proprietary data formats, and making analysis tools accessible to non-coding scientists

  • Core tech stack uses Python with libraries like scikit-image, Dash for web interfaces, and Q-Pi for GPU acceleration. Focus on balancing commercial needs with open source tools

  • Data analysis pipeline includes acquisition, reconstruction, alignment, segmentation and reporting - automated through web interfaces that simplify complex workflows

  • Team uses multiple approaches for deployment including VMs, servers near instruments, and cloud platforms depending on specific needs

  • Development happens through a mix of in-house work by core data science team and academic partnerships/sponsorships, especially through Oxford

  • Key focus areas include improving pore network structures, nanoparticle analysis, and atomic-level imaging to understand catalyst performance

  • Web platform allows scientists to upload data, run analysis, and generate reports without coding while preserving complex analysis capabilities

  • Emphasis on making tools performant and user-friendly while handling large datasets through techniques like Dask for parallelization

  • Future goals include enabling real-time analysis and expanding capabilities for live microscopy data processing