AI ENGINEER · GENAI
I'm an AI intern at Ideen and a Software Engineering student at UTFPR. I build agents, MCP servers and RAG pipelines that run in production, with tests, gates and security from the first commit.
Claude Code skills that take a PEDIDO.md to a running system, and a script, not the agent, approves every stage.
5/5 requests with 100% of acceptance criteria met, 4 of them unseen
A single MCP server with per-user permissions and auditing that became the front door to the company's agent tools.
560+ automated tests in the gateway
From a Streamlit prototype to an ADK agent with A2A on Agent Engine: every architecture change answered a real problem.
7 architectures iterated over 7 months
Every serious project starts with a spec and a plan versioned in the repo, including this site.
See exampleIn Fábrica, acceptance tests are sealed by hash before any code. The MCP gateway has more than 560.
See exampleModel output is untrusted input: structural gates, format validation and no critical rule living only in a prompt.
See exampleInfrastructure as code, CI/CD, Kubernetes and security in other teams' repositories.
Connected Cloud Build to GitHub at the organization level and unblocked an internal app deploy pipeline with least-privilege permissions.
Migrated deploy and tests of an internal app from GitHub Actions to Cloud Build with remote state, and updated spec and README to match reality.
With the site in private mode, the root returned 302, the GKE health check marked the pod unhealthy and the gateway served 503 everywhere. A HealthCheckPolicy on an always-200 endpoint fixed it.
Merged an AI-generated version with a secure one: Argon2id, AES-256-GCM for personal data, rate limiting, four-role RBAC, bootstrap-only admin and no made-up KPIs on the dashboard.
If your team builds with LLMs, I'd love to help.
Message me on LinkedIn