Selected work

Built properly. Explained simply.

A selection of systems our engineers have designed and built – each summarised as the challenge, what was built and the result.

Discuss your project

Security, Bot & Fraud Protection

Ad-traffic bot filter

The challenge: A paid advertising campaign was losing budget to bots, click farms and scrapers that never bought anything.
What was built: A high-speed filter in front of the landing page that checks every click and stops bots and data-centre traffic before they reach the offer – with automatically updated blocklists and a live traffic report.
The result: Only real visitors reached the offer page, and the advertiser could finally see how much of their traffic was fake.

Cloudflare
Edge filtering
Automated blocklists
Traffic reporting

AI Agents & Automation

AI support agent on AWS

The challenge: A support team spent hours answering policy questions and handling routine refunds by hand.
What was built: An AI agent on Amazon Bedrock that reads company documents, checks orders in the database, works out refund eligibility, processes refunds and opens support tickets – with guardrails that automatically hide card numbers and personal data.
The result: A complete, tested loop from question to action, deployed to the cloud with infrastructure as code.

Amazon Bedrock
RAG
OpenSearch Serverless
AWS Lambda
EventBridge
Terraform

DevOps & CI/CD

Streaming web app with automated releases

The challenge: A streaming web application needed safe, repeatable releases and a clear view of how it was running.
What was built: A full CI/CD pipeline: Docker containers, Jenkins builds with SonarQube code-quality and Trivy security scans, GitOps deployment to Kubernetes with Argo CD, and Prometheus and Grafana dashboards.
The result: Every change is built, scanned and released automatically, with live dashboards showing the health of the app.

Docker
Jenkins
SonarQube
Trivy
Kubernetes
Argo CD
Prometheus
Grafana

Cloud Cost Optimisation

Cloud cost scanner

The challenge: Teams were paying every month for AWS resources nobody remembered creating.
What was built: A scanner that checks every AWS region in seconds using read-only access – finding unattached volumes, idle IP addresses, unused NAT gateways, empty load balancers, old snapshots and logs that never expire.
The result: A clear list of waste and what each item costs per month – without any risk to live systems.

Python
Boto3
AWS
Read-only IAM

Security, Bot & Fraud Protection

High-speed edge traffic router

The challenge: A busy website needed strong bot protection without slowing down real visitors.
What was built: An in-memory traffic router combining Cloudflare edge filtering with application-level checks – rate limiting, hidden honeypots and browser-hint validation – plus background blocklist updates that need no restarts.
The result: Strong protection with no noticeable delay for real visitors, and continuous automated checks of security headers and site health.

Cloudflare
Rate limiting
Honeypots
Security headers
Synthetic monitoring

Security, Bot & Fraud Protection

Real-time payment fraud detection

The challenge: Online payments needed fraud checks without the cost of enterprise tools.
What was built: A machine-learning fraud scoring system on AWS Lambda and DynamoDB that checks each transaction against multiple fraud indicators through a real-time API, with an interactive dashboard.
The result: Suspicious transactions flagged as they happen, on a cost-effective serverless set-up.

Machine learning
AWS Lambda
Amazon DynamoDB
Amazon S3
Real-time API

Cloud & Hosting

Secure serverless configuration export

The challenge: Settings stored in AWS needed to be exported safely as part of automated deployments.
What was built: A custom AWS Lambda function, deployed with CloudFormation, that reads values from Parameter Store, writes them to S3, logs to CloudWatch and uses only the minimum permissions it needs.
The result: Reliable, repeatable deployments – including clean removal, so deployments never hang.

AWS Lambda
CloudFormation
Parameter Store
Amazon S3
CloudWatch
IAM

AI Agents & Automation

AI social media automation

The challenge: Posting consistently on LinkedIn took hours every week.
What was built: An n8n workflow that writes posts with GPT-4, creates matching images with DALL-E and publishes them on a schedule.
The result: A steady, consistent social presence without the daily manual effort.

n8n
GPT-4
DALL-E
Scheduling

Examples of systems our engineers have designed and built, including internal and proof-of-concept projects. Details are anonymised and no client names are shown.

Have a similar challenge?

Tell us about it – we'll explain the options in plain English.

Book a free consultationExplore services Call Us