Earth Observation Data Pipelines on AWS

Why this exists now

Amazon SageMaker geospatial capabilities entered maintenance mode on 30 July 2026. Organisations doing earth observation work on AWS now build the pipeline themselves, on Amazon S3, AWS Batch, AWS Lambda, STAC and Cloud Optimised GeoTIFFs.

That is ordinary data engineering wearing a geospatial hat, and it is work most teams can specify but few have time to build.

The position this is for

There is imagery. It arrives from a supplier, or from the free Sentinel and Landsat archives, or from an aircraft, and it lands in a bucket. Someone wrote a script. The script works on their machine. Processing a season takes a week and nobody is quite sure it produced the same answer as last time.

The analysts want a time series. What exists is a folder.

Assessment

One imagery estate. You receive:

• An inventory of what you hold, in what format, at what resolution, over what period, and how much of it is actually addressable.
• Source analysis: what is free and what is licensed, what the licence permits you to publish, and where each source sits relative to the AWS region you process in.
• A target architecture: catalogue, storage tiering, processing and serving, with the cost of each at your volume rather than at a reference volume.
• Residency position, naming the region for every stage, which matters when the imagery is of your own country and the processing is not.
• A reproducibility plan, so the same inputs produce the same outputs and you can prove which version of which model produced a given result.

Build

We build it in your account: ingestion from your suppliers and from the open archives, conversion to Cloud Optimised GeoTIFF, a STAC catalogue so the imagery is searchable by time, place and property rather than by filename, processing on AWS Batch and AWS Lambda that scales with the job instead of with the budget, S3 lifecycle policies so cold imagery costs what cold imagery should, and tile serving through CloudFront for the people who just want to look at it.

Where analysis is in scope we build the models alongside it, including change detection and classification, with Amazon Bedrock where the output needs to be read by a person rather than a machine.

Acceptance is a measured comparison: the same scene processed through the new pipeline and the old one, agreeing to an agreed tolerance, at a measured cost and runtime.

Operations

Monthly, if you want it. Scheduled acquisition and processing, catalogue maintenance as new sources arrive, model version management with accuracy measured after each change, storage cost review as the archive grows, incident response, and a monthly report covering volume processed, cost per scene and pipeline reliability.

Support

Enquiries through this listing are answered within two business days, with a scoping call and a written proposal. After purchase you get a named engagement lead, weekly status reporting, and support during New Zealand business hours. Response targets by severity are in the statement of work attached to each offer.

Documentation

Service descriptions and case studies: https://www.easycoder.co 

Infrastructure costs

Everything runs in your AWS account. AWS charges for S3 storage and requests, Batch and Lambda compute, CloudFront, and data transfer are billed by AWS directly to you, separately from the price of this engagement. Imagery licensing remains between you and your supplier. Our access is through IAM roles with least-privilege policies that we document and hand over.