Senior Data Engineer · Insurance data · Platform builder
I turn messy, multi-source data into pipelines, contracts and products that hold up in production. At Hiscox I build and modernise insurance data integrations across six European markets. As founder of Aragón Tecnologia I own a portfolio of SaaS products end to end, from architecture to the cloud bill, with a squad of AI agents I supervise.
const renato = {
role: "Senior Data Engineer",
focus: "reliable data platforms",
cloud: ["Azure", "AWS"],
agents: "supervised, scored, accountable",
approach: "evidence over assumptions",
scope: "architecture to operations"
}// about
I work across data integration, migration, modelling, orchestration, and platform reliability. A green pipeline is not proof to me: I trust a release when its output reconciles row by row, its contract is explicit, and the decision behind it is written down. When something breaks, I go to the root cause and protect what is shared.
I also bring a product owner's perspective. Through Aragón Tecnologia, I take products from idea to production and remain accountable for what happens afterwards. That operational responsibility shapes how I design data systems and how I lead technical work.
// results
What the platforms changed for the people using them.
The first five at AG Capital Group, the last four at Hiscox. A capacity run then took the modernised flow to 50,000 rows.
// stack
// experience
Insurance data integration for Europe: bordereaux from brokers and coverholders in Ireland, France, Germany, Spain and the Benelux, validated, transformed and routed to policy systems on Azure Databricks, Data Factory and SQL Server.
Enterprise batch pipelines for large-scale automotive and operational data, in a regulated, international environment.
Led the evolution of a central analytics and operational platform: 14+ integrated systems, 8 business units and 324 daily users.
Data platforms and automation for clients in Brazil and Portugal, and a portfolio of SaaS products built and run end to end.
education Postgraduate in Big Data & Analytics, UNISUL (in progress, 2024–2026) · Business Administration studies, UDESC
languages Portuguese (native) · English (C1) · Spanish (intermediate)
// recent
Operations work behind the products: cheaper infrastructure, accountable AI agents, and monitoring that watches itself.
Moved a multi-app production fleet from x86 to ARM64 (AWS Graviton) with no data loss. Each app was deployed and verified on the new platform before any traffic moved. CI stays on x64 and cross-packages the ARM builds of Prisma engines and native modules (sharp, canvas). Everything is defined in Terraform, and secrets move between hosts end-to-end encrypted. The AWS run-rate is about 60% lower.
A voice-driven chief of staff that supervises the AI coding agents working across my repositories. Each agent gets a weekly scorecard (delivery, rigour, reliability, autonomy) and a critical review. Two in-house agents keep watch on cloud cost and the security posture. A holographic HUD shows the whole squad orbiting the core in real time.
Uptime, database health and TLS expiry are checked every five minutes, from outside the cloud they monitor, with alerts only when a state changes. A GitHub Actions watchdog confirms the monitor is alive, and the monitor confirms the watchdog in turn, so if the monitor goes silent, that silence is itself an alert.
// work
Vertical SaaS through Aragón Tecnologia software and data architecture with applied AI, built for trust, security, and precision.
Clinical management platform for psychologists records, scheduling and finance in one place. 21 validated clinical scales with automatic scoring, plus AI that drafts records, flags dropout risk, and surfaces progress insights.
Digital platform for nutritionists pairing AI-assisted nutritional analysis with real-time patient tracking and teleconsultation cutting analysis time and improving adherence through a digital food diary.
Management platform for beauty salons and studios scheduling, automated WhatsApp confirmations, and real-time financial reporting that cuts no-shows and helps independent professionals grow.
Client onboarding without the email chase. Workspaces, projects and invited clients share one guided flow, so every document and task has a place and an owner.
Jobs in Portugal for international talent: candidate profiles with documents, job search and alerts that bring matching roles to candidates instead of waiting for them.
Full portfolio at aragontecnologia.com →
// open source
Public data engineering projects: real patterns, tested and CI-checked, built in small reviewed pull requests. Synthetic data only.
Real-time retail events through Kafka (KRaft) and Spark Structured Streaming: windowed aggregations with watermarks, a dead-letter queue, an Iceberg sink with exactly-once semantics, and CDC with Postgres and Debezium.
Production-shaped batch ETL with PySpark: incremental loads over a high-watermark, partitioned outputs, quality gates with volume anomaly detection, run summaries for observability, and architecture decision records.
Analytics engineering with dbt and DuckDB: staging to marts with four layers of data tests, from column contracts and referential integrity to cross-layer revenue reconciliation. Runs end to end at zero cost.
Reusable Terraform module for an AWS data lake on S3, Glue and Athena: secure by default, plan-time input validation that mirrors the real API constraints, and fmt, validate and tflint gates in CI.
Apache Airflow orchestrating a retail medallion pipeline (bronze, silver, gold) with a data-quality gate, retries and idempotent backfills. Runs locally, no cloud.
A medallion lakehouse on Delta Lake feeding an MLflow-tracked churn model, over synthetic SaaS product-analytics data. Runs locally, no cloud.
Kanban flow metrics on a live event stream: throughput, cycle time, WIP and percentile-based delivery forecasting with Kafka and PySpark Structured Streaming.
Maps heterogeneous sources (CSV, JSON, SQL, Parquet) into a canonical model and runs business rules over it, splitting valid rows from rejects with every violated rule reported per row.
Natural language to SQL with an LLM: the model proposes, a deterministic guard disposes. A minimal lexer normalizes each query and blocks anything that is not a single read-only SELECT.
A real product in the open: next-train departures and trip planning for Portugal on CP's open GTFS data. Daily sync pipeline with a quality gate, service-calendar aware planning, pricing from CP's public API, and a dbt analytics layer.
Self-hosted portal for families in an estate process: shared documents, automatic income splitting and individual statements. A dbt + DuckDB analytics layer sits on top, with a reconciliation gate that fails CI if a single cent goes unaccounted for. Synthetic data only.
More on github.com/renatoaragon →
// contact
Based in Porto and open to remote data engineering and architecture roles across the EU.