Powering real-time AI analysis with scalable, cost-saving AWS infrastructure
About the client
MotionMiners GmbH, based in Dortmund, Germany, is developing technology to optimize manual processes in logistics and production environments. Its integrated mobile sensors and custom machine learning algorithms collect employee movement and location data and highlight processes that can be improved by identifying inefficiencies and ergonomic strain.
The challenge
Our client came to us with the AI component of their project — wearable mobile sensors and custom AI algorithms to analyze the data. What they lacked, however, was the infrastructure for processing the sensor data at scale using their AI algorithms in order to show data at user dashboards. This cloud-based environment had to be:
Delivered value
Solution
While the MotionMiners team defined the overall environment’s architecture, our engineers focused on bringing it to life. As Lemberg Solutions is an AWS-certified partner, the project was supported by the AWS Incremental Workloads Program, allowing MotionMiners to accelerate development.
One of the most significant parts of the project was orchestrating the flow of data — the infrastructure had to trigger AI analysis when the data appeared in the cloud. To achieve this, we built the entire infrastructure around an event-driven architecture, where every new sensor data automatically triggers the next stage of the workflow. Overall, the system includes two main application layers: one responsible for initial data pre-processing and another for AI analysis and delivering final results to user dashboards.
After sensors upload data to the cloud infrastructure, AWS Lambda, a computing service, initiates a processing workflow. Then, the event-driven autoscaler, KEDA, monitors incoming workloads and triggers Karpenter, a tool for provisioning GPU resources. Moreover, the architecture supports multi-region deployments, so users can process data in their target region.
Once processing is complete, the infrastructure scales back down to its basic settings. If additional data is still waiting in the queue, compute resources remain active to avoid wasting time scaling the system up again.
To ensure the reliability of data flow, our engineers implemented the OpenTelemetry tool, which gives our client full visibility into the data flow in the pipeline. With monitoring dashboards in place, they can quickly detect stalled processes, prevent data loss, and verify that information successfully passes through every stage of the pipeline.
Infrastructure security was ensured by separating it into production and management environments to mitigate operational risks and enable easier rollback to previous settings. The communication between services is restricted to private cloud networks, so no traffic is exposed to the public network.
The сollaboration with Lemberg and the AWS Incremental Workloads Program helped us accelerate our development tremendously. Not only on a technical level but also knowing that we could rely on them. We had a strict deadline for this project and were able to get going in a matter of days and be production-ready within the quarter. An achievement that would otherwise not have been possible.
How it works
