Raw data into decisions.
AI into systems that actually ship.

I'm André Figueiredo — Data Engineer working across dbt, Microsoft Fabric and Databricks, and building production-grade AI agents on top of governed data. This site is a showcase of what I build and how I build it.

Belo Horizonte, Brazil · Data Engineering · DBT · Fabric · Databricks · Azure · AI Engineering · Agents · RAG

// 01 — expertise

Data Engineering

dbt

Dashboards nobody trusts? That's exactly what dbt fixes for me: tested, documented analytics layers — every metric with an owner, a test, and lineage.

Microsoft Fabric

Data scattered across silos? I unify it all in Microsoft Fabric — from OneLake to Power BI — into a platform the business team can actually use.

Databricks

Pipelines breaking at 3am? I design lakehouses on Databricks with medallion architecture, Delta Lake, and orchestration that warns before it breaks.

Azure

Pipelines with no versioning or deploy pipeline? I orchestrate ingestion and ETL in Azure Data Factory and put it all under CI/CD in Azure DevOps — pipelines that are tested, reviewed, and shipped without manual deploys.

// 02 — ai engineering

AI Engineering

Most AI projects die at the demo. Mine start with what everyone else leaves for later: real architecture, retrieval over governed data, validation, and metrics.

★ FEATUREDworkshop-agent-harness

A sales team drowning in five years of data needed answers, not another dashboard. I built an agent that talks to the database (text-to-SQL over PostgreSQL), retrieves qualitative context (RAG over Qdrant), and delivers action-ready sales-improvement reports — backed by a full engineering harness: scoped rules, validation hooks, MCP integration, subagents, and metrics.

User → LangGraph Agent → [ text-to-SQL → PostgreSQL | RAG → Qdrant ] → Report → FastAPI → React
View repository ↗

Agentic systems

Multi-step agents with scoped tools, memory, and guardrails, not single prompts.

RAG & retrieval

Retrieval pipelines grounded in governed, versioned data — not just a vector store.

Evaluation & reliability

Metrics, validation hooks, and test harnesses that catch regressions before users do.

// 03 — projects

Projects

AI ENG★ FEATURED

workshop-agent-harness

A sales team drowning in five years of data needed answers, not another dashboard. I built an agent that talks to the database (text-to-SQL over PostgreSQL), retrieves qualitative context (RAG over Qdrant), and delivers action-ready sales-improvement reports — backed by a full engineering harness: scoped rules, validation hooks, MCP integration, subagents, and metrics.

View repository ↗
DATA ENGIN PROGRESS

dbt-analytics-project

Modular modeling + tests + docs over a public dataset.

DATA ENGIN PROGRESS

fabric-lakehouse-demo

End-to-end Fabric lakehouse + Power BI semantic model.

// 04 — about

About

I'm a Data Engineer based in Belo Horizonte, Brazil, working across dbt, Microsoft Fabric, and Databricks — building analytics layers and lakehouses that teams can actually trust and query with confidence.

What pulled me toward AI engineering was simple: agents and RAG systems are only as good as the data underneath them. So I build the platform and the agent together — governed data feeding retrieval and reasoning that actually holds up in production.

Same discipline either way — whether the artifact is a SQL model or an LLM agent: tested, documented, versioned, and built to be trusted, not demoed once and forgotten.

$ data_stack
  • · dbt
  • · Microsoft Fabric
  • · Databricks
  • · Delta Lake
  • · Spark
  • · SQL
  • · Power BI
$ ai_stack
  • · LangGraph
  • · RAG
  • · Qdrant
  • · PostgreSQL (text-to-SQL)
  • · FastAPI
  • · React
  • · MCP
// 05 — contact

Let's talk data & AI.

Always open to trading ideas on data platforms, AI engineering, and how to make the two work together — find me on GitHub or LinkedIn.