Enriched at the door
One POST from the reef to Amazon API Gateway and AWS Lambda. Amazon DynamoDB stamps site, location and organisation onto every event.
Case studyProject OceanOasis
From the reef to the dashboard.
How Project OceanOasis (PO2) gets species detections from cameras on remote reefs into live, multi-tenant dashboards on AWS, without a single server to manage.
- Industry
- Marine monitoring, conservation tech
- We built
- A serverless ocean data platform
- Runs on
- AWS Lambda, AWS Glue, Amazon Athena
4 units
Live, in 2 countries
Camera units sending today
<1 s
p95 ingestion, zero errors
Load test · 50 requests a minute
500
Concurrent users, no intervention
Stress test
The challenge
The best sites have no infrastructure.
Ocean monitoring still runs on divers, clipboards and sporadic surveys. The best sites, from remote reefs to quiet coves, have no fibre, no power grid and no backend nearby, so by the time surveys are processed, conditions have moved on.
A single detection is unreliable. A thousand of them, time-stamped and placed, paint a picture of our ocean.
Project OceanOasis
On LinkedIn, September 2026
What we built
Serverless, from reef to dashboard.
Assembly designed and built the proof of concept in three phases, ingestion, a medallion data lake and multi-tenant analytics, on managed AWS services with near-zero idle cost.
What we delivered
A medallion data lake
AWS Glue turns raw JSON in Amazon S3 into one Parquet row per detection, then a Gold table partitioned by organisation and date.
Queryable on arrival
Amazon Athena reads the lake in place. Partition projection makes new data queryable the moment it lands, with no crawlers.
One dashboard, a view per org
Amazon QuickSight with row-level security: each organisation sees only its own detections, mapped over its own sites.
01 / 04
Built in
Least privilege, end to end
No AWS credentials on the reef: each unit holds an API key that can post to one endpoint. Scoped IAM roles in the middle, row-level security at the dashboard, media served from private buckets through Amazon CloudFront, and Amazon CloudWatch and AWS CloudTrail keep it in view.
- AWS IAM
- Amazon CloudFront
- Amazon CloudWatch
- AWS CloudTrail
Powered by AWS
From reef to dashboard.
- Sent from the reef. Cameras detect species on the device and send compact JSON events, not video, over satellite to Amazon API Gateway, authenticated with an API key.
- Enriched on arrival. AWS Lambda stamps each event with fleet, site and organisation data from Amazon DynamoDB, writes lean JSON to Bronze and moves images, logs and config to Silver.
- Structured in the lake. AWS Glue jobs turn each pack into one Parquet row per detection and build the Gold fact table, partitioned by organisation and date.
- Queried and shown. Amazon Athena queries Gold in place. Amazon QuickSight applies row-level security per organisation, and Amazon CloudFront serves images from private buckets.
- Governed throughout. AWS IAM keeps every hop least-privilege, and Amazon CloudWatch and AWS CloudTrail log and audit it.
High-level architecture
Project OceanOasis on AWS
Key results
Live, tested and ready to scale.
Live
4 units
Sending from 2 countries
Detections land enriched, partitioned and queryable, from the reef to the dashboard.
Load test
<1 s
p95 at normal load
Real device packs at 50 requests a minute for 10 minutes, with zero errors.
Stress test
500
Concurrent users
The serverless path scaled on its own, well above what the current fleet sends.
Turning single detections into that picture is cloud work […] When we say a detection reaches shore in a usable format, this is who we mean.
Project OceanOasis
On LinkedIn, September 2026
Our certified expertise
AWS Advanced Tier Services Partner and Well-Architected Partner Program. All 12 current AWS Certifications are held by our team.