Agent runtime engineering
Engineering a production agent harness with Pi
Embedding the Pi SDK in a governed runtime for operational investigation, code review, feedback, and model evaluation.
Pi-based production harness
Read case studyFrom data to working intelligence.
AI & Data Engineer
I build AI agents, data platforms, and the systems that connect them.
AI agents & evaluation
Data platforms & pipelines
Cloud & infrastructure
Brasília, Brazil · Building since 2018
Engineering experience across
- AI agents
- Software
- Data platforms
- Machine learning
- Cloud
Selected work
Complex problems. Working systems.
A closer look at the architecture, decisions, and delivery behind my consulting and in-house engineering work.
All workDurable software delivery
Smart Agent Factory: durable delivery for AI-built software
A custom graph and execution runtime that turns a reviewed request into isolated implementation, real-data validation, and an evidence-backed draft pull request.
Custom durable agent runtime
Read case studyMulti-agent orchestration
Building a verified multi-agent delivery platform
Turning data-platform requests into isolated, auditable delivery workflows with specialized agents, human gates, and independent verification.
Artifact-gated agent delivery
Read case studyData platform engineering
Scaling a data platform on Kubernetes
Migrating Spark workloads from Dataproc to a Kubernetes platform while standardizing ingestion, orchestration, and Data Lake access.
Data platform modernization
Read case studyEach case study distinguishes my contributions from the work shared with engineering teams.
- lower monthly BigQuery cost
- 25%Query and storage optimization
An outcome behind the architecture
Cloud efficiency, grounded in the details.
BigQuery optimization, grounded in query patterns, table design, and workload analysis.
Read the optimization caseWays I can help
Where engineering meets your next challenge.
From a workflow that needs automation to a platform that needs a new foundation. Start with the problem; choose the tools around it.
Let's talk about your challengeAI agents & automation
Put AI into a real workflow.
Connect agents to your tools and data, with evaluation, execution controls, and human review built into the workflow.
Explore agent deliveryData engineering
Make your data useful.
Connect ingestion, orchestration, and data-quality practices so your team can build on a shared, usable data foundation.
Explore data platform workCloud & platform engineering
Build a stronger platform.
Bring software, cloud infrastructure, and delivery workflows together, with clear interfaces and systems your team can inspect and operate.
Explore platform engineeringThe engineer behind the work
Hi, I'm Matheus.
I work across AI, data, and infrastructure because a useful system rarely stops at one discipline.
My experience spans retail data platforms, machine learning infrastructure, and agents and execution tooling.
More about my approachPublic projects
Code you can inspect.
Agent workflows, Spark orchestration, and change data capture. Public repositories with implementation notes.
All projectsGoal design for autonomous coding loops
Give autonomous coding agents a clear goal and a reliable way to know when they are done. Includes templates and worked examples.
- Agent engineering
- Claude Code
- Goal design
Spark on Kubernetes orchestration with Airflow
Run Spark jobs on Kubernetes from Airflow, with a custom operator, working deployment examples, and Google Cloud Storage integration.
- Apache Airflow
- Apache Spark
- Kubernetes
- GCS
Change Data Capture on Kubernetes
Stream MySQL changes through Kafka and Debezium on Kubernetes. Includes connector configuration and a FastAPI consumer you can run.
- CDC
- Kafka Connect
- Debezium
- Kubernetes
Writing
Notes on AI agents, engineering, and data.
Articles on agent loops, specification-driven development, and data platform architecture.
All writingJul 13, 2026 / Matheus Jericó on Medium
Stop Prompting Claude Code. Write the Loop That Prompts It.
A concrete design for a standing loop that finds work, builds context, dispatches an agent, verifies the result, records progress, and decides what happens next.
Mar 23, 2026 / Matheus Jericó on Medium
AI Needs Specs, Not Just Prompts
An argument for treating specification as the central engineering constraint in AI-assisted development, with intent, architecture, and verification kept in sync.
Dec 08, 2025 / Data Engineer Things on Medium
From Delta Lake to Apache Iceberg: A Complete Migration Blueprint for Modern Data Platforms
A migration blueprint for teams moving from Delta Lake toward an open, multi-engine Iceberg architecture, including tradeoffs, operating concerns, and post-migration practices.
Your next engineering challenge
What are you building?
Bring the problem, the constraints, and where you want to go.
Let's discuss how I can help you get there.