From monolith to microservices in biotech research 

Our team supported the client in building and growing the platform that combined cloud engineering, cheminformatics knowledge, AI expertise and modern software development. The aim was not just to deliver new tools but to create a connected system that could support daily scientific work at scale. 

We first set up the cloud environment on AWS and adapted it so the platform could run reliably at scale. Over more than two years, we continued to optimise performance, control costs and improve workflows on AWS. Later, when the client decided to switch cloud providers, we carried out a smooth migration to Google Cloud Platform. This ensured continuity while also giving better scalability and clearer cost control. 

We created custom Kubernetes tools and cost monitoring solutions to keep the system efficient and affordable. A new retrosynthesis engine was developed in house and linked directly with automated laboratories and reagent stocks. 

On the AI side, our MLOps engineers built and maintained scalable infrastructure to support machine learning. This included automation for training, deploying and serving models, using Anyscale (Ray) and Weights & Biases for experiment tracking. AI and ML were applied in areas such as predicting compound properties and generative chemistry, which supported the automated design–make–test–learn cycle in drug discovery. 

We also designed backend services for AI inference that could process millions of requests reliably, and used Snowflake and cloud data lakes to combine and manage complex scientific data. On top of this, we developed web applications with React and GraphQL, giving scientists clear and simple interfaces without the need for deep cloud knowledge. 

Through these efforts, the client gained a platform that was easier to use, more reliable and better suited to the scale of modern drug discovery.