Cloud Infrastructure for Techem’s Digital Boiler Room

Background and Goals

Existing heating systems often waste a lot of energy because they are not operated at optimal settings. Our customer Techem retrofits heating systems with IoT-connected sensors to capture time series data and apply machine learning. This results in recommendations and warnings delivered via a customer portal and push notifications. Such actionable advice improves the efficiency of a heating system by up to 15%, saving money and reducing CO2 emissions.

Techem was looking for a team that can design, build and maintain the data stores and backend APIs to handle a scaled roll-out into thousands of buildings while keeping the processing latency at near real-time levels.

Our Solution

Our Data Engineering Team was brought in to design and implement a scaled-out version of an existing prototype that happened to be in production already. We went with an evolutionary architecture approach, that gradually transitions the prototype into a maintainable production-ready implementation.

For this “brownfield” integration, the data platform must manage multiple IoT-Events-Streams and correlate them with a common Master Data Layer. Optimized data products are maintained for consumption by Data Science Modules as well as the Frontend Portal.

Case Img

Highlights

A complex product development initiative involving domain matter experts, IoT hardware specialists, Data Scientists and Software Engineers successfully delivers a product that is being rolled out into thousands of buildings. The agile setup allows teams within the program to iterate fast on customer feedback.

Technologies

Analytical workload on Azure ADX. Azure SQL DB as a transactional store. Ingestion via Azure IOT Hub. Master Data from GraphQL API backed by Azure CosmosDB.
REST-based API Backend on AKS using JAVA Spring Boot Data Science on Azure Function Apps using Python. Authentication and IAM based on Azure AD B2C.
React-based SPA for Frontend incl. React Flow visuals. Testing with Cypress. Observability with Azure Monitor. Deployment Automation with Gitlab, Helm, Terraform