Every service has a fixed scope, deliverables and price per tier, agreed in writing before work starts. The cycle: audit, design, build, maintain and train.
A read-only review of an AWS account: where money is being wasted, what security risks there are and what to fix first, with the estimated savings of each change. Nothing is touched.
A fast website with no servers to patch, in your own AWS account and fully described in code. Certificate, CDN, firewall, automated deployment and properly configured email, with fixed scope and price.
A prioritised report of what should change and why.
Access, MFA, permissions, public exposure, logging, backups and alerts, checked against AWS and CIS best practice.
Oversized instances, unnecessary NAT Gateways, orphaned volumes, logs kept forever. A report with the estimated saving of each change.
Pipelines, secrets, permissions, run times and deployment approach, with an improvement plan.
Current state against what the business needs: risks and recommendations in a high-level design document.
Which tools are used, what the agents are allowed to do and what data leaves. Report and draft policy.
High- and low-level documents, justified decisions and an estimated monthly cost.
Static website, serverless application, containers on ECS, Kubernetes on EKS or multi-account with guardrails.
Design on Amazon Bedrock: sources, permissions, cost and risks. No self-hosted models.
Everything in Terraform, documented, with role-based access and never shared keys.
Everything that was clicked together in the console brought into Terraform: imports, modules, remote state and documentation. Optional: a profile with the technical controls of Spain's ENS.
Locked-down root, MFA, Identity Center instead of IAM users, CloudTrail, GuardDuty and budget alerts.
Automated deployments with GitHub Actions, GitLab CI or AWS, with environments, rollback and OIDC access, no stored keys.
From an unpatched VPS or flaky hosting to something stable, with automatic TLS, backups and predictable cost.
An automated backup policy and a real restore, timed and documented.
Alerts that mean something, uptime checks and centralised logs without paying a fortune.
Tidy DNS and properly configured email: SPF, DKIM and DMARC.
An architecture document generated from the infrastructure code, with the design decisions written up.
The report's findings applied in Terraform and documented, with support hours included.
Automations and reports: cost alerts, backups, clean-up. Code and usage documented.
Claude Code and similar tools set up per repository: project rules, agent permissions, connections to your tools (MCP) and safe handling of secrets.
Automated review of every pull request with your own rules: security, bugs and style.
Let the AI query tickets, documentation or inventory with read-only permissions.
Building the designed assistant: ingestion, search, permissions and cost control.
In writing, asynchronously, with no 24/7 on-call.
Updates, cost review and written support, with fixed hours each month.
Hosting, deployments, updates and cloud cost for the agency's clients, white-label if preferred.
A monthly report with alerts and recommendations, compared with the previous month.
Prepaid hours for written changes and questions, with an expiry date.
Best practice changes with every release and every AI model: periodic review of what is installed.
Hands-on sessions on your own account or code.
One-to-one or team sessions on AWS, Terraform and CI/CD.
Claude Code and similar tools day to day: context, rules, agents, review and security, with exercises on your own code.
What AI does well and badly, which data must not leave, and a usage policy. Helps meet the training obligation of the EU AI Act.
Practical fundamentals on your own account or code.
One-off help for anyone who wants to get something running or understand how it works, in plain language and with a fixed quote before starting.
Tell me what you're running and what's worrying you. I'll get back to you within two working days.