🏗 System Architecture

Machine Sentinel emits telemetry data from its internal PLC (programmable logic controller), which is processed by local software running directly on the machine. This software collects relevant performance and event data, formats it, and uploads it to a cloud-based ECS service.

The ECS service acts as the central cloud bridge and performs the following actions:

• Writes incoming data to S3 and CSV archives

• Pushes structured metrics to Kinesis streams

• Logs transactional data in DynamoDB

• Sends machine-level metrics to a custom Amazon service called DataOnboarder via an API called UploadData

The UploadData API ensures data is correctly routed and registered across internal analytics pipelines managed by DataOnboarder.

There are multiple dashboard sites that consume data from different sources:

• 📦 PackScout Dashboards – A site consisting of dashboards visualizing pack success rate, failure causes, and trend analytics. Data originates exclusively from substreams routed through DataOnboarder.

• 📊 RoboDetective Dashboards – An analytics-focused dashboard suite that queries inventory-type data directly from AWS DynamoDB tables.

• 🌡️ MachinePulse Dashboards – Displays temperature and sensor data, reading directly from a Kinesis stream and historical snapshots in S3.

These dashboards vary in latency requirements and data formats, and rely on the data flowing correctly through each tier.