Cloudfic

Cloudfic

Consulting in security, development, automation and monitoring.

We take systems that grew in layers, with hand-made changes and failures nobody sees, and leave them with infrastructure as code, business logic in a repository you own, and alerts where there used to be silence.

Architecture from one of our cases, AI agents for IT support in a lab:

Teams and WhatsApp messages reach a router that hands them to each client SDK; the SDK classifies with an LLM and operates the ticketing system, Zabbix and the HR ERP RPA.Teams, WhatsAppmessagingRouterone portSDK per clientworkers, TypeScriptITSMticketsLLMclassificationZabbixL3 agentRPA, ERPHR
Teams and WhatsApp messages reach a router that hands them to each client SDK; the SDK classifies with an LLM and operates the ticketing system, Zabbix and the HR ERP RPA.

Services

Each service starts from a concrete problem. The result we show is a measured number, with where it was measured.

Security

Problem: Clusters and cloud accounts changed by hand pile up excess permissions and unreviewed configuration, and the gaps only show up at audit time.

How we work

  • Terraform as the only change path, with tfsec and SonarCloud in CI.
  • Least-privilege IAM and CI authenticated over OIDC, with no long-lived keys.
  • Read-only AI audit agents that look for leaked keys and Terraform drift.

Measured result

30%

fewer PCI audit gaps in Kubernetes clusters

Measured at: our technical lead's previous employer, global retail

Development

Problem: Business logic gets locked inside a vendor's low-code tool and is lost at every migration, and deploys depend on manual steps.

How we work

  • TypeScript SDKs, with business logic in a repository you own and out of the vendor's low-code.
  • CI/CD with zero-downtime deploys and environments provisioned from code.
  • Testing through the real channel before every delivery, and a suite that runs without real credentials.

Measured result

60%

shorter deploy time, with fewer manual errors

Measured at: our technical lead's previous employer, financial SaaS

Automation

Problem: Flows that grew in layers, with no versioning or logging, duplicate records, keep processing finished sequences and fail without warning.

How we work

  • Idempotency and concurrency guards in the orchestration.
  • A state snapshot before every change.
  • LLM classification with a deterministic fallback, and every agent decision logged.

Measured result

Fewer than 40

records in the follow-up queue, down from about 1,300

Measured at: consulting client, B2C solar energy

Monitoring

Problem: Integrations that stop with no alert and channels that are configured but send nothing: the incident surfaces when a user complains.

How we work

  • Observability with Grafana, Prometheus, Dynatrace and Zabbix.
  • A full map of the system before the first change, including the queues and channels that fail silently.
  • Cost dashboards and budget alerts.

Measured result

40%

shorter incident resolution time

Measured at: our technical lead's previous employer, global retail

Cases

Told from the client's side: context, problem, what we did, result, and the system diagram.

  1. Lead qualification and follow-up for a solar energy company

    Consulting client, B2C solar energy

    We audited and stabilized a B2C solar energy company's lead automation: the follow-up queue dropped from about 1,300 records to fewer than 40.

    Workflow automation platform, Conversational agent, Cloud CRM, Relational database, Async queue

  2. Moving a multi-tenant SaaS to another cloud

    Consulting client, B2B SaaS

    We migrated all three environments of a schema-per-tenant B2B SaaS off hand-built VMs without interrupting active tenants, with an estimated payback under two months.

    MySQL, PHP, Reverse proxy, Object storage, Managed database

  3. AI agents for IT support, in a lab

    IT support provider, lab and proof of concept

    For an IT support provider, we moved the service assistant off a low-code platform into in-house TypeScript SDKs integrated with the ticketing system, as a proof of concept in a lab.

    TypeScript, Node.js, LLM classification, ITSM, Microsoft Teams

  4. From low-code flows to an AI-orchestrated lead pipeline

    Consulting projects

    We built lead qualification pipelines (WhatsApp capture, AI qualification, CRM sync) in n8n, Jitterbit and SkyOne, and now solve the same problem with an agent that decides each step.

    Claude, Cursor SDK, n8n, Jitterbit, SkyOne

Technical lead

Iago Bonifácio leads every project. He is a specialist cloud engineer with eleven years on GCP, AWS and Azure, across global retail, fintech, multi-cloud consulting and SaaS.

Iago's experience and certifications

Contact

For a project or a technical assessment, reach us by email or LinkedIn.

iagorbonifacio@gmail.com
OG-AI email
firstogai@og-ai.app
LinkedIn
linkedin.com/in/iago-r-g-bonifacio
GitHub
github.com/og-ai-app