[ Case study · Enterprise ]
Knose pet insurance.
AI-powered clinical-history summaries, data-lake analytics, automated banking integration, and a custom insurance risk-management system — engineered on AWS for an Australian pet insurer.
- Client
- Knose · knose.com.au
- Sector
- Pet insurance, Australia
- Role
- Senior full-stack engineer
- Stack
- Node.js · AWS · AI/ML
> summarize clinical history — done
> load analytics lake — done
> reconcile banking run — done ✓
[ What was built ]
Four workstreams, one platform.
Each piece shipped as its own increment and plugged into the existing insurance platform.
- 01
AI-powered clinical-history summaries
Veterinary records arrive as long, unstructured histories. The summarizer turns them into a readable, consistent brief for the people assessing a pet's cover and claims.
How
An AI summarization pipeline with structured output, so every summary follows the same shape and can be reviewed and corrected by a human.
- 02
Data-lake analytics
Operational data from across the platform lands in one place that analysts can actually query, instead of living inside individual services.
How
An AWS data lake fed by the platform's services, with analytics built on top for reporting across policies, claims, and operations.
- 03
Automated banking integration
Money movement that used to need manual handling runs on its own, with the reconciliation trail an insurer needs.
How
Serverless integrations with the banking provider, event-driven and idempotent, so retries never double-post.
- 04
Custom insurance risk-management system
A purpose-built system for assessing and managing risk on the book, shaped to how the underwriting team works rather than a generic template.
How
Node.js services on AWS serverless with the rules and workflows the team defined, integrated with the rest of the platform.
[ Engineering choices ]
What made it hold up.
The same choices carry into every enterprise engagement I take on.
Human in the loop where it counts
AI drafts the summary; a person makes the decision. The system is built so that review is quick and correction is easy.
Serverless by default
AWS Lambda and event-driven services keep the platform cheap when quiet and elastic when busy, which insurance workloads are.
Boring, auditable money paths
Banking automation is designed for reconciliation first: every transfer traceable, every retry safe.
[ More work ]
Unified pension platform · Singlife · Collab-Cart · OpenLumin
[ Next step ]
Building something similar?
A free 30-minute technical call. Bring the constraints — I'll tell you what I'd build first.
● Typically responds within 24–48 hours