mateus.bonn

Backend & Distributed Systems · Brazil · Open to remote

I build event-driven backend systems that stay up when things go wrong.

Backend engineer working on the JVM — distributed messaging, resilient service design, and clean architecture. Currently building production-grade pipelines with Java 21, Spring Boot, and Apache Kafka.

See my work GitHub available for new projects
// event-driven rca pipeline log → kafka → analysis → store
Architecture of an event-driven root-cause-analysis pipeline An incoming log POST is published to a Kafka topic, consumed by a Spring AI worker, analyzed by an LLM, and stored in PostgreSQL. A dashed branch shows retry, circuit breaker, and dead-letter-queue handling. POST /logs ingest Kafka log-events Worker Spring AI LLM Bedrock · Ollama PostgreSQL rca store retry · circuit breaker · DLQ

How I work

I treat reliability as a feature, not an afterthought.

Most of the interesting work in backend systems happens at the edges — what happens when a message is redelivered, when a downstream call fails, or when you need to swap a provider without touching business logic. These are the defaults I build with.

01

Resilient by default

Circuit breakers, retries with exponential backoff, and dead-letter routing — so a single bad event degrades gracefully instead of taking the service down.

02

Idempotent & safe to retry

Deduplication with SHA-256 hashing means a redelivered message is processed exactly once. At-least-once delivery stops being a liability.

03

Provider-agnostic design

One interface, swappable implementations. Switch an LLM between local Ollama and AWS Bedrock through a build profile — zero code changes, zero lock-in.

Tools

The stack I reach for.

Grouped by what they do, not by buzzword count. Highlighted items are where I'm strongest.

Language & Frameworks

Java 21 Spring Boot 3.4 Spring AI JPA / Hibernate Maven

Messaging & Streaming

Apache Kafka Event-driven design Dead Letter Queues Idempotent consumers

Resilience & Patterns

Resilience4j Circuit Breaker Retry / Backoff SOLID

Data, Cloud & AI

PostgreSQL AWS Bedrock Ollama Docker Compose Git / GitHub

Selected work

Projects, with the decisions behind them.

I care less about the feature list and more about why each piece is built the way it is.

Event-driven · Java 21 · Kafka

Root Cause Analysis Pipeline

A backend service that ingests application logs, routes them through Kafka, and uses an LLM to produce structured root-cause analyses — designed end-to-end for resilience and token-cost control.

  • Thin Kafka boundary delegating to a focused analysis service — one interface, one implementation, clear seams.
  • SHA-256 idempotency so redelivered events never double-process or double-spend on LLM calls.
  • Resilience4j circuit breaker around the LLM, with backoff retries and DLQ routing for poison messages.
  • Stacktrace truncation to cap token cost before each model call.
  • Provider switch between local Ollama and AWS Bedrock via Maven profiles — no code change.
View source ↗

Built with

Java 21 Spring Boot 3.4 Spring AI Apache Kafka Resilience4j PostgreSQL AWS Bedrock Docker
Your next project

Project title goes here

One or two sentences on what it does and who it's for. Replace this card with a real project, or remove it.

  • A standout technical decision or trade-off you made.
  • A result or constraint you handled well.
View source ↗

// PLACEHOLDER — replace with a real project

Built with

Tech Tech Tech

Contact

Let's build something reliable.

Open to remote backend and distributed-systems roles. The fastest way to reach me is email or LinkedIn.

Email me LinkedIn GitHub