ArquitechCorp · Cloud architecture and integration

Before writing code, the decision must be recorded.

We architect cloud, integration, and AI governance for operations that cannot stop. Every engagement starts with an evidence-based diagnostic — not a proposal of hours — and ends with the decision documented, versioned, and auditable.

View services Schedule a 30-minute conversation

Since 2013 Cloud architecture and integration across operations in Brazil, Angola, and South Africa.

100% of our infrastructure is code Decisions in ADRs, operations in runbooks, environments provisioned by IaC.

Data origin in Brazil This site is served from Magalu Cloud, br-ne1 region, behind Cloudflare.

The problem

Three symptoms, almost always the same cause.

The cause is almost never a missing tool. It is a missing recorded decision: someone chose a path at some point, and nobody knows why. Without it, every incident becomes archaeology and every project becomes a rebuild.

Integrations without ownership

The failure point is between systems, inside neither of them. It costs more than the server, and it fails less often than the average suggests — but it fails at the worst possible moment.

No API catalog and no interface contract
Silent errors: a 200 response carrying the wrong data
Nobody tracks who called what, when, or how much

Cloud cost without visibility

The bill arrives consolidated, with no attribution per product or team. Without tagging, budget, and an owner, there is no optimization — only cuts made in the dark.

Orphan resources and forgotten environments
No chargeback: nobody answers for the spend
Scaling by the most expensive path, not the simplest

AI without containment

The demo works. Production stalls at the committee — or ships without scoping, audit trail, or emergency brake.

Personal data in prompts with no declared legal basis
No decision log: the result cannot be explained
No kill switch: the agent keeps acting in the dark

Services

Architecture engagements with declared scope and a named deliverable.

Every engagement is sold as a defined scope of consulting work, with a named deliverable and acceptance criteria. Where it makes sense, the initial diagnostic is a fixed, paid project — which avoids starting a commercial relationship with a blind estimate.

Cloud architecture and landing zone

A cloud foundation designed for audit: identity, network, policy, data territory, and cost — before the first production workload.

Systems integration and APIs

A governed integration layer with versioned contracts and end-to-end observability — so the failure surfaces before the customer complains.

Data and AI readiness

Before buying a model, check whether the data can support the answer. We assess lineage, quality, classification, and access — and we say when the data is not ready.

AI agent containment and governance

A pilot reaches production only when the boundary is known: scope, authority, audit trail, emergency brake, and incident response.

View details and deliverables

Cases

Anonymized cases, with what was measured.

For contractual confidentiality, no case is published with a client name, identifiable sector, or commercial value. Each case carries the context, what was done, and — where it exists — the indicator measured. Cases marked as illustrative are technical designs drawn from real problems, still without attributed production measurement.

Illustrative

Industrial group · multi-country integration

An integration platform replacing dozens of direct connections

Context: Operations in three countries with local systems talking to each other over direct connections. Every new partner required a new project.
What was done: A centralized integration layer with versioned API contracts, messaging for the asynchronous flow, and end-to-end tracing with business-impact alerting.
Result: Onboarding a new partner stopped being a project and became configuration, over an interface that already existed and was already monitored.

Illustrative

Retail · data product

Data readiness before spending on AI

Context: Board pressure to launch an AI use case on a short deadline, over a data environment with no documented lineage.
What was done: A readiness diagnostic: source inventory, lineage, quality criteria with an owner per domain, sensitivity classification, and access control.
Result: The pilot was sliced to start with the domain that already had trustworthy data, and the two blockers that would have invalidated the demonstration became visible before the model was purchased.

In production

ArquitechCorp infrastructure · institutional site

Brazilian cloud as a verifiable argument, not a promise

Context: A large part of the market’s offering for data residency in Brazil exists only in a contract and in sales material.
What was done: Static content served from Object Storage at Magalu Cloud, br-ne1 region, with Cloudflare in front for DNS, TLS, and cache, and a Worker handling routing, cache policy, and security headers.
Result: The claim is verifiable at any moment: the site loads nothing from a third party, and the response originates in Brazilian territory. The same holds for the governance discipline — the decisions behind this environment are public.

In definition

Brazilian enterprise · search risk and brand exposure

Brand exposure treated as risk, with a register and an owner

Context: A significant part of what is said about a large company today is produced by search engines, aggregators, and language models — and is governed by nobody inside it.
What was done: A proprietary framework that reads search exposure as business and compliance risk, with a risk register, authority matrix, responsibility model, and executive scoreboard.
Result: Under construction — this is the firm’s next product. The first measured case will be generated from the operation described in the item above, and published once there are numbers.

How we work

Four steps, and none of them starts with a quote of hours.

The method is the same at any engagement size. What changes is the depth — and the right to say, at the end of step two, that the project should not be done.

Evidence-based diagnostic

Before suggesting any change, we measure the current state: real application response, integration latency, consumption and cost attribution, data quality, and exposure surface. The deliverable is a finding with proof, not a diagnostic of perception.
Recorded decision

Every architectural choice becomes a short, dated record: what the context was, what was decided, what was rejected and why, and which risk was consciously accepted. That is what lets the architecture be defended two years later, when whoever decided has already left.
Implementation with the client's team

We execute alongside the team, not in its place. The goal is to leave the environment and, above all, the ability to maintain it. Environment provisioned by code, change through a pipeline, operations in a runbook — so delivery does not depend on our presence.
Ongoing operations and evidence

A metric with an owner, alerts driven by business impact, and a periodic review of what was decided. When an assumption changes, the record changes — with a date and a reason. Architecture without review is just debt with good looks.

Who we are

Track record before credential. Method before résumé.

ArquitechCorp is a cloud architecture and integration practice. The work is led by the person who signs the design — and the portfolio behind this page is made of public decisions, not vendor logos.

Thiago Ribeiro · Cloud & Integration Solutions Architect

Since 2013 in cloud and integration architecture across large-scale operations in Brazil, Angola, and South Africa. Focus on the integration layer, cloud governance, automation through infrastructure as code, and containment of AI agents in enterprise environments.

Current certifications: Microsoft Certified: Azure Fundamentals (AZ-900) · Microsoft Certified Solutions Associate (MCSA, legacy) · Acronis (technical certification) · Databricks Fundamentals.
Framework alignment: TOGAF — aligned by practice, not certified. We state it that way because an uncertified credential is not a credential.
Working languages: fluent in PT-BR and English.

Sovereign hosting as evidence, not as infrastructure

This site is served from Magalu Cloud Object Storage (br-ne1 region) with Cloudflare for DNS, TLS, and cache. No third-party resources are loaded — no webfont, no CDN script, no tracking pixel. That is why the CSP can be genuinely restrictive and why this page needs no cookie consent banner.

$ curl -sI https://arquitechcorp.com/en/ | grep -i "x-amz-storage-class"

Describe the problem. We will tell you if it is an architecture problem.

Schedule a 30-minute conversation. No institutional presentation — bring the symptom (integration degradation, a cloud bill, a stalled AI pilot) and we leave the call with an initial technical hypothesis and what would need to be measured to confirm it.

Schedule a conversation

No form in this version. Direct channel, response within one business day.