Case Study

Forecasting Climate Risk for 1,500 Automotive Plants Worldwide

A serverless AWS analytics pipeline that scores production plants against 8+ natural hazards and translates each one into financial exposure for OEM customers.

Overview

Turning Hazard Data into Plant-Level Financial Risk

The client, a global ratings and market analytics provider, wanted to give its automotive and OEM customers a view of physical risk at each production plant. The product had to predict hazards such as coastal flooding, wildfire, drought, water stress, extreme heat and cold, and cyclones, and show what each would cost.

The Challenge

The data came from multiple sources held in separate AWS accounts, which made this a distributed computing problem from the start. Workloads fluctuated sharply, so compute had to scale without overspending. Multi-step batch jobs depended on one another, and a single failure could not be allowed to cascade through the pipeline.

The hard part wasn't the models, it was getting data from every source to run reliably every time. JRD broke the problem into pieces we could trust, and our customers now get plant-level risk they can act on.

Anonymous Head of Climate and Sustainability Analytics, global ratings provider

Retries and failure handling were designed in before the first model ran. A failed job now stops cleanly and recovers on its own instead of taking the whole nightly run down.

Anonymous Cloud Engineering Lead, global ratings provider
What We Delivered

Four workstreams, one risk pipeline

Cross-Account Data Integration

  • Connected data sources across multiple AWS accounts into one processing framework
  • Broke the pipeline into small functional units that could be built and tested independently
  • Integrated outputs with the client's existing enterprise systems

Event-Driven Orchestration

  • Triggered and chained workflows with cloud events and AWS Step Functions
  • Built retry and failure handling to stop errors cascading across batch jobs
  • Managed dependencies between multi-step jobs on AWS Batch and Fargate

Cost-Aware Serverless Compute

  • Ran variable workloads on Lambda, Fargate and Batch, sized per job
  • Balanced instance types for cost against processing time
  • Tracked spend with AWS Budgets and Cost Management

Resilience and Reporting

  • Encrypted data at rest and in transit, with fine-grained IAM access
  • Configured cross-region replication and automated backups
  • Delivered risk reporting to analysts through Power BI
The Result

Plant-Level Risk. Reliable Runs. Controlled Cost.

The client's OEM customers can now see which plants face which hazards, and what that exposure could cost. The pipeline scales with demand on serverless compute and recovers from job failures without manual restarts. It runs across regions with automated backups, so a regional outage does not stop the product.

8+Natural hazard types modeled
1,500Plants assessed
35%Lower compute cost vs. baseline
Let's Talk

See what reliable risk analytics could do for your customers

Tell us where your data lives and what has to run every night. We'll come back with a practical plan and a scoped first step.