Cold Storage Analytics – Lightweight ECS-Based Processing Platform (NDA)

For a client in the cold storage & refrigeration industry, we designed and implemented a
lightweight, containerized analytics platform to process specialized weekend-only operational data.
The objective was to enable accurate AI-driven performance analysis while keeping infrastructure lean and cost-efficient.

Requirements

• Develop a minimal yet robust Ruby-based application capable of running as Docker containers in AWS ECS.
• Implement two core containerized processes:
  • Dispatcher: Cron-driven job scheduler that places processing tasks into SQS queues.
  • Worker: Consumers that pull tasks from SQS, process them, and store results.
• Restrict processing to only two specific time slots:
  Saturday 11:59 PM and Sunday 11:59 PM — ensuring data is analyzed only when assets (cold rooms, deep freezers) remain untouched.
• Integrate with InfluxDB for both raw and processed data storage.
• Invoke a custom AI model to perform deep asset analysis and store results for dashboard visualization.

What We Delivered

Technical Implementation

• Developed a custom Ruby application with a lightweight, Rack-based internal framework to keep the codebase fast and modular.
• Containerized both dispatcher and worker services using Docker for ECS deployment.
• Implemented a cron-based dispatcher that schedules one job per asset per day to the SQS queue during the defined time slots.
• Created a worker process that:
  • Pulls a job from SQS.
  • Fetches the corresponding day’s asset data from InfluxDB.
  • Sends the data to the AI model for analysis.
  • Stores the AI-generated JSON analytics back into InfluxDB.
• Ensured the analytics results were readily available for the client’s dashboard without performance bottlenecks.

Outcome

• Lightweight, cost-effective architecture optimized for AWS ECS and containerized workloads.
• Accurate, clean AI analysis by leveraging weekend-only untouched asset data.
• Scalable SQS-based job orchestration capable of handling hundreds of assets in parallel.
• Fast data turnaround enabling near-immediate updates to the dashboard analytics.

Client name withheld under NDA. The above describes technical architecture and delivery approach only.