Senior Data Engineer · Software Builder · Founder @ Aragón Tecnologia
Cloud data platforms on AWS by day. Shipping AI-driven SaaS products on the side.

• 🛠️ Senior Data Engineer building large-scale data pipelines and platforms on AWS.
• 🧱 Founder & engineer at Aragón Tecnologia, where I take products from idea to production: architecture, development, deployment and operations.
• 🤖 I integrate AI where it genuinely improves a product, not as a gimmick.
• 🌍 Based in Porto, Portugal. Working in international, English-speaking teams.
• 💬 Ask me about data engineering, AWS, PySpark, infrastructure as code, and building SaaS solo.

Data Engineering

Architecture & Patterns

AWS & Cloud

Software & Web

Tools are the disposable part. The framework I reach for today will be replaced sooner than I would like, and that is fine. What lasts is the architecture: a design that is sustainable, scalable and durable rarely has to change, whatever tool implements it underneath. So I optimize for the shape, not the stack.

The projects below are laid out along the path data takes through a platform. Each is a decision rather than a demo, runs on synthetic data, is verified in CI, and carries a short Design principles note that explains the why. Read together, they are one opinion about how a data platform should be built.

      INGEST              PROCESS            ORCHESTRATE          STORE + MODEL           CONSUME
 ┌────────────────┐ ┌────────────────┐ ┌──────────────────┐ ┌──────────────────┐ ┌──────────────────┐
 │ kafka-retail-  │ │ spark-retail-  │ │ airflow-retail-  │ │ delta-lakehouse- │ │ dbt-duckdb-      │
 │ streaming      │ │ etl            │ │ orchestration    │ │ mlflow           │ │ analytics        │
 │ stream + CDC   │ │ batch ETL      │ │ medallion DAG    │ │ Delta + MLflow   │ │ nl-to-sql        │
 └────────────────┘ └────────────────┘ └──────────────────┘ └──────────────────┘ └──────────────────┘

 foundations    terraform-aws-datalake (IaC)  ·  data-mapping-framework (canonical model)  ·  kanban-streaming-metrics (flow metrics)
 in every repo  explicit schemas  ·  data-quality gates  ·  idempotent, replayable design  ·  tests + CI

Small, focused data-engineering projects. Each one runs with synthetic data and is verified in CI.

Project	What it demonstrates	CI
kafka-retail-streaming	Kafka (KRaft) + Spark Structured Streaming: event-time windows with watermarks, dead-letter queue, Iceberg sink with exactly-once semantics (ADR), CDC with Postgres + Debezium merged via foreachBatch
spark-retail-etl	PySpark batch ETL (raw → clean → curated): incremental loads over a high-watermark, partitioned outputs, quality gates with volume anomaly detection, run summaries, ADRs
airflow-retail-orchestration	Airflow TaskFlow DAG orchestrating a medallion pipeline: quality-gate branching, dynamic task mapping, idempotent partitioned writes, DAG-integrity and end-to-end tests
delta-lakehouse-mlflow	Medallion lakehouse on Delta Lake with an MLflow-tracked churn model: idempotent silver via MERGE, full-recompute gold, tracked and registerable modelling
kanban-streaming-metrics	Streaming flow metrics with Redpanda + PySpark: throughput, cycle time, WIP, and percentile-based forecasting (p85/p95, SLA attainment)
dbt-duckdb-analytics	Analytics engineering with dbt + DuckDB: layered staging/marts and four layers of data tests, up to cross-layer revenue reconciliation
terraform-aws-datalake	Reusable Terraform module for an AWS data lake (S3 + Glue + Athena): secure by default, plan-time input validation, tflint gate
data-mapping-framework	Map multiple platforms/formats (CSV, JSON, SQL, Parquet) onto one canonical model, with business rules and full violation reporting per rejected row
nl-to-sql	Natural-language-to-SQL with Claude: a deterministic read-only guard with lexer-based query normalization

Real products, code in the open — not demos: these run in production with real users.

Beyond the open-source projects above, I build and run a small fleet of production SaaS products at Aragón Tecnologia, mostly on my own, from the first commit to the 3am alert when something breaks.

How I run them:

• Each one is multi-tenant SaaS on AWS, released with zero-downtime blue-green deploys: a small Node agent flips nginx over once the health checks pass.
• I operate them, not just ship them: self-hosted uptime monitoring, error tracking with Sentry, structured logging, scheduled jobs, and automated LGPD retention.
• Some are realtime, using Server-Sent Events for live dashboards and WebRTC for video calls.
• I pull data across products into a medallion lakehouse (bronze, silver, gold) on Databricks with an MLflow model on top, fed by anonymized exports.
• And I add AI where it genuinely helps the person using the product, never for show.

The products and their source stay private. What you see in this profile are the same engineering patterns, rebuilt from scratch with synthetic data.

Open to business partnerships and collaboration on data platforms and products. Reach me at aragontecnologia.com or on LinkedIn.