Client Success Story

Turning a Jupyter Notebook Model into a Licensable Commercial Product 

Yahara Software transformed SentrySciences' bench-tested machine learning model — used to characterize subvisible particles in biologic drug products — from research code in Jupyter notebooks into a fully deployable, licensable commercial software product. 

• The Challenge

  SentrySciences had already solved the hard scientific problem of developing a complex neural network AI model that characterizes subvisible particles and protein aggregates in biologic drug products that generates "fingerprint" models used for formulation development, scale-up, and production quality control. But the model lived in a Jupyter notebooks and needed to be scaled for commercial use.
  

• How Yahara Helped

  Yahara took the model from prototype to commercial product:

  • Rebuilt the research code as a standalone Python library — a fully source-controlled, versioned, maintainable codebase in place of scattered notebook cells
  • Wrapped it with a real user interface, making the model's core capabilities — creating a fingerprint model, comparing new images against a fingerprint, curating particle images — usable through software rather than raw code
  • Packaged the result for distribution, engineering it to be deployable, maintainable over time, and licensable as a standalone commercial product

  
  

   

• The Results

  • A commercially available product, ParticleSentryᴬᴵ, compatible with flow-imaging microscopy, backgrounded membrane imaging, and microfluidic imaging platforms

  • A maintained, versioned codebase replacing what had been research-only notebook code, ready for ongoing updates rather than one-off runs
  • The ability to retrain fingerprint models as new conditions are captured, keeping the product useful as customers' science evolves
  • A model now used by biopharma companies, generating license revenue from work that previously lived entirely on the bench

• Why It Matters

  For any organization whose AI breakthrough is currently stuck in a notebook, our work for SentrySciences is proof of what it looks like on the other side.

• FAQ: What does it take to turn a research ML model into a commercial product?

  For SentrySciences, it meant rebuilding notebook-based research code into a standalone, source-controlled Python library, wrapping it in a real user interface, and packaging it so it could be deployed, maintained, and licensed — the same path Yahara calls AI Model Integration.