The challenge
The platform publishes digital catalogues and personalised offers for local shopping to millions of users. Behind it, retailers' orders and advertising campaigns were set up, validated and delivered through steps that still involved manual handling. At that scale, every manual step is a delay and a source of inconsistency, and it caps how many campaigns the team can deliver.
Separately, the AWS footprint had grown with the product and nobody owned the bill.
Approach
The order and campaign flows were rebuilt as pipelines: events on queues, stateless workers, and services that validate and deliver campaigns without a person copying data between systems. Work was delivered inside an 8-person cross-functional Scrum team, test-first, with Ahsan leading test design and mentoring junior engineers in the client-facing platform area.
The cost initiative was run company-wide: measure what each service actually uses, right-size, remove idle resources and codify the infrastructure so it does not drift back.
Architecture
- User-facing REST services in Python (Flask, FastAPI) and Java (Spring Boot) in a microservices architecture.
- AWS ECS and Lambda for compute, SQS for decoupling pipeline stages, S3, SES, Cognito and CloudFront around them.
- Infrastructure as code with Terraform and CloudFormation.
- Redis and Elasticsearch for caching and search, Docker and Kubernetes for packaging and orchestration.
- Observability with Grafana and a test-driven workflow with PyTest.
Results
Automating the order and campaign pipelines increased order volume and campaign delivery efficiency by 15 to 20 percent. The cost initiative reduced AWS spend by 15 percent across the company.
This work was delivered by Ahsan Sheraz as an engineer on the platform's team. It is included here because it is the same pipeline-first approach to automation that ILMOTECH applies to client systems.
- 14M+Users served by the platform
- 15 to 20%Increase in order volume and campaign delivery efficiency
- 15%Reduction in AWS costs, company-wide