Digital Transformation in Research and Development for a leading Chemical Manufacturer
Reinforcing an organization’s collective resolve to accept and adapt to digital tools requires a bottom-up transformation roadmap – one that starts by forward-looking digitization of smaller processes culminating in a complete, yet seamless digital transformation of a business function.
The leadership team of a chemical manufacturing giant partnered with Mu Sigma for a similar digital transformation. Through our strategic partnership, we are helping them overcome the inertia of sticking to legacy architectures and are enabling a shift towards establishing a shared digital and analytics capability.
The leading chemical manufacturer wanted to optimize the time spent on fulfilling requests for their most common experiment – Thin Layer Chromatography. These requests help researchers understand chemical composition, troubleshoot underlying impurities, identify causes of miscoloration in products, and so on.
The research and development team had been storing images and reports collected over the last 20 years in different locations across various formats. This data was not made widely accessible and hence was barely used.
As a result, researchers did not leverage the historical experiment data (images of thin layer chromatography plates and customer reports) and had to start afresh for every new experiment request.
It took considerable time collating information from past chemical experiments to fulfill retrieval requests.
The problem ran deeper, so we broke it down into brief problem statements:
• Research data, especially images, was stored in an unstructured manner
• Research processes lacked standard naming conventions
• There was missing and duplicate data
• The legacy infrastructure did not have search capabilities
• Lack of a collaboration platform to document and share best practices
This called for a digital transformation solution that reduced search time and standardized research data storage for faster access and more accurate results.
The Mu Sigma Approach
Digital transformation of the process required a clear comprehension of the standard operating procedure for these experiments. We collaborated with their research and development analytical team to decode every step that needed to be digitized.
We envisioned a cloud-based tool that would allow the research and development group to perform a free text search on reports and image search. It would help in phasing out the legacy system in the following ways:
• An image processing algorithm to extract features from images of widely different varieties, quickly and efficiently
• A robotic Python automation script to convert legacy image data to a consumable format
• Standardized reports and image naming conventions to enable quick retrieval
Of these, identifying the image processing algorithm posed a unique challenge. The variation in image quality made it impossible to use a single algorithm to extract relevant image features.
Fortunately, we have always approached digital transformation empowered by a low-cost extreme experimentation capability. With iterative experimentation throughout the solution-design phase, we arrived at the differentiating features of the solution. In this case, low-cost experiments helped us devise an ensemble algorithm – a combination of multiple techniques – to drive highly efficient and effective feature extraction from images.
Leveraging this approach, the Mu Sigma team created and implemented a powerful thin layer chromatography digital database web app capable of:
• An image search to find similar reference plates of experimental conditions from a database of nearly 36,000 images
• A cognitive search that taps into historical research knowledge and customer reports
• A reference plate search to identify compounds directly from thin layer chromatography reference plates
The tool was hosted as a cloud-based web-app, with the entire tool architecture built on the Google Cloud Platform (GCP). The choice of GCP makes the app highly scalable across research and Development centers (in different geographies) and other techniques.
This framework has enabled some key functionalities:
• Significant time savings for researchers drawing insights from the historical data
• Elastic text-based search across the reports
• Advanced image-based search for the historical research data
This research catalyst has transformed the way research analyses are conducted for the client in a research-intensive industry.
Using this tool, the client has realized:
• 50% time/cost saving by avoiding certain downstream analyses
• ~5 hrs saved per request for thin layer chromatography analyses performed by new researchers
More importantly, it has empowered new researchers who have limited experience to upskill themselves and perform experiments at a practitioner level.
Click here to know about how Mu Sigma helped the R&D department of a leading Pharma company to accelerate research.