AI solutions architecture · Python & TypeScript · Founder of Kite
I'm a developer from Brazil working on AI agents, integrations, and developer tools. I like being involved in both the product and the code: understanding the business rules, deciding how the pieces should fit together, and implementing the solution.
My background spans product, technical delivery, customer operations, and infrastructure support, including Hostinger and TCS in Hungary. At VanguardIA, I managed 16 concurrent B2B technology programs and delivered 25+ automations and integrations into production alongside engineering and client teams.
That experience keeps me close to what happens after deployment: whether the workflow is useful, what breaks when an integration fails, and how the team will operate it.
Customer Success logic defined in code: activation, lifecycle, health, and risk. Product events feed into versioned TypeScript definitions, and the engine computes each customer's state with the rule context behind it.
I'm building Kite around a CLI, a typed configuration DSL, and a Node.js SDK. It reflects something I care about in architecture: business rules should be explicit, reviewable, testable, and explainable.
An AI customer-service case study for a fictional restaurant, combining chat, voice, orders, reservations, CRM synchronization, and human handoff.
The project separates model interpretation from business operations: CrewAI routes requests, while application code validates pricing, availability, and confirmations. Azure AI Foundry provides model access, and Azure audio services support transcription and live voice. Operators can take over a conversation with its context intact.
Built with: TypeScript, Next.js, React, Python, FastAPI, CrewAI, Azure AI Foundry, SQLite, OAuth 2.0, and RD Station CRM. A Three.js restaurant provides an interactive entry point.
| Area | Technologies |
|---|---|
| Languages & backend | Python, TypeScript, FastAPI, Node.js |
| Cloud & AI platforms | Microsoft Azure, Azure AI Foundry |
| Agents & orchestration | CrewAI, LangGraph, LLMs |
| AI observability & evaluation | PostHog AI Observability, Langfuse, DeepEval |
| Data | PostgreSQL, SQL, Supabase, SQLite |
| Integrations & automation | REST APIs, OAuth 2.0, webhooks, n8n, Make |
| Web & interactive experiences | React, Next.js, Three.js |
| Infrastructure | Docker, Linux, VPS, SSH |
- Start with the process, business rules, and the people who will use the system.
- Make boundaries clear between model responses, application state, and external services.
- Keep human handoff and failure handling part of the workflow.
- Use prototypes to resolve technical uncertainty and inform the roadmap.
- Treat evaluation, observability, and documentation as part of building the solution.
Alongside building Kite, I'm open to AI solutions architecture, AI engineering, and technical leadership opportunities, as well as collaborations around agents, integrations, and developer tools.
Based in Brazil. I work in Portuguese and English.