Renato Garcia

Software Architect · Distributed Systems & Applied AI

I have spent 20 years getting systems to work together, first with ESB, then Kafka, and now AI agents. Today I run the architecture behind 30+ marketplace integrations and an AI platform that answers about 4,000 real-estate leads a month.

Londrina, Brazil

Java Kotlin Quarkus Apache Kafka Kubernetes Spring AI LLM Agents

Read the work

20+ Years in software engineering

30+ Marketplace integrations live

1.3k Requests per second in production

4k Leads a month handled by AI agents

About

Most of my career has been about integration: getting systems that were never meant to talk to each other to exchange data reliably and at volume. I started with ESB and SOA middleware, moved to Kafka and event-driven microservices, and in the last couple of years to AI agents. I joined GuBee when it started and designed its platform, which connects e-commerce stores to more than 30 Brazilian marketplaces and handles about 1,300 requests per second over some 170 services. I also built the engineering team there. In 2025 I started INAI at Integra.do, another company in the GuBee group. It is an AI platform for real-estate agencies, in production since January and answering around 4,000 leads a month.

Technical direction: architecture standards, large migrations (cloud, Java 25, GraalVM native, OpenTelemetry) and the roadmap for them

AI in production: multi-agent systems, RAG, guardrails and LLM cost tracking with Spring AI and Koog

Hands-on: Java, Kotlin, Quarkus, Spring Boot, Kafka and Kubernetes, in e-commerce, telecom, banking, credit and the public sector. Built and lead a team of up to 9 engineers

20+ Years of experience

30+ Marketplace channels

1.3k Requests per second

4k AI-handled leads a month

Experience

Founding Software Architect

Integra.do (INAI) GuBee group · AI platform for real estate · Remote

Current · concurrent with GuBee

Jul 2025 – Present

Designed and built INAI from scratch. Its agents reply to real-estate leads on WhatsApp at any hour, qualify them and pass the good ones to brokers. Imobiliária Inglaterra has used it since January 2026, with about 4,000 leads a month
Split the work among five agents (sales, billing, support, maintenance and new launches) with a router between them, which kept each agent small enough to test properly
Built the WhatsApp and portal intake (ZAP, VivaReal, OLX, ImovelWeb) so leads are not lost or duplicated when something fails
Picked a model size per task and put cost and latency per client on dashboards, which kept the LLM bill under control
Wrote about 70% of the 2,400+ commits plus the architecture and roadmap docs; the rest of the team works on top of those standards

Kotlin Spring Boot 4 Spring AI PostgreSQL + pgvector Redis Streams Kubernetes

Context

INAI is Integra.do’s AI platform for real-estate companies. Its agent persona, Bruna, works as a digital SDR on WhatsApp: it qualifies leads coming from property portals and chat, searches and recommends listings, and hands qualified leads to brokers. Integra.do is part of the GuBee group. Imobiliária Inglaterra was the first client and has run on it since January 2026, with about 4,000 leads a month. I founded the codebase, CI and Kubernetes setup, wrote the AI core, and drive the roadmap that turns it into a multi-tenant SaaS.

Responsibilities

Own the system architecture: hexagonal Kotlin services, multi-tenant data model, agent runtime and integration adapters
Design the agent core: hybrid router (keyword + semantic + session affinity, LLM tie-break), agent-to-agent handoff and per-tenant overrides
Build lossless ingestion: WhatsApp on Redis Streams (consumer groups, pending-entry recovery, per-tenant debounce) and ZAP, VivaReal, OLX and ImovelWeb leads with dedupe and replay protection
Design retrieval: hybrid search (pgvector + SQL + full text fused with RRF) with a local ONNX cross-encoder reranker, and geo ranking on PostGIS and OpenStreetMap POIs
Build the agent runtime: dynamic tools (HTTP, sandboxed GraalJS scripts, MCP), SpEL rules, Handlebars prompt templates, three-layer memory, semantic LLM cache and two-tier guardrails
Control LLM cost and latency by tiering four models by task (chat, classification, insights, transcription) and tracking spend per tenant
Write the platform roadmap (admin and RBAC, financing simulation, CRM connectors, SSO, developer portal and SDKs, voice) and sync it with Jira

Stack

Kotlin 2.3 Java 25 Spring Boot 4 Spring AI 2 OpenAI MCP PostgreSQL pgvector PostGIS Redis Streams WhatsApp Cloud API OpenTelemetry Grafana Kubernetes GitLab CI React

Impact

Went from an empty repository to the first client in production in about six months
Led a hardening milestone: tenant isolation on 43 controllers, BCrypt API keys migrated with a dual-read window, about 200 blocking calls removed, oversized modules split
About 3,200 automated tests cover the agent flows, routing and intake
11 Grafana dashboards make LLM cost, latency, prompt-cache hit ratio and the re-engagement funnel visible per tenant

Senior Staff Engineer

GuBee Full-time · Londrina, PR

Current role

Apr 2019 – Present

Designed and still own the architecture of the integration hub between e-commerce platforms and 30+ marketplaces. It runs on Kafka and handles about 1,300 requests per second across roughly 170 services
Leading the move from Huawei Cloud to Tencent Cloud to lower infrastructure cost: about 170 services, 36 MongoDB databases and 25 Kafka consumer groups, switched over one by one with feature flags and a rollback plan for each
Moved services to Quarkus native, which cut memory use per service by more than 60%, and added OpenTelemetry tracing and logs across the platform
Compared four agent frameworks and chose JetBrains Koog for our AI products: a WhatsApp sales assistant that goes all the way to PIX payment, and automatic catalog enrichment. I wrote up the decision in “I had ruled out Koog”
Wrote 15 Claude Code plugins with our migration, scaffolding and troubleshooting steps, so other engineers can do that work without waiting for me
Hired and lead the engineering team (up to 9 people) and made hexagonal architecture, DDD and TDD our default

Java 25 Kotlin Quarkus Apache Kafka Kubernetes OpenTelemetry

Context

GuBee is a SaaS integration hub that connects e-commerce platforms (VTEX, Shopify, WooCommerce, Magento) to the major Brazilian marketplaces (Mercado Livre, Amazon, Magalu, Shopee and 30+ others). I joined when the company started, designed the system from scratch and have kept working on it since, together with the team I hired.

Responsibilities

Design and maintain the microservices architecture handling millions of daily product, order and stock synchronisation events
Define and enforce engineering standards across all services: Hexagonal Architecture, Domain-Driven Design, Test-Driven Development
Lead the migration from Spring Boot to Quarkus native, reducing memory footprint and cold-start times
Architect event-driven pipelines on Apache Kafka for real-time marketplace synchronisation
Mentor developers, own the technical backlog and drive product decisions with the business
Run the infrastructure: Kubernetes, Kafka, Redis and MongoDB clusters, first on AWS and now on Tencent Cloud
Write the migration playbooks (strategy, service checklists, secrets, storage, cutover order) that let several engineers migrate services in parallel
Ran the framework POCs (Koog, Embabel, LangChain4j, Spring AI). Koog went to production after its 1.0 release fixed the Spring Boot 4 incompatibility
Architect the AI checkout as a thin BFF: commerce rules stay in the core engine (enforced by ArchUnit), while the BFF owns conversation, guardrails, LGPD consent and the WhatsApp 24-hour window; quality is gated by DeepEval suites for faithfulness, refusal and prompt injection
Design catalog enrichment with confidence scoring and human review, semantic search on OpenSearch kNN, and Kafka-driven processing
Codify recurring engineering work as AI agent skills: new marketplace scaffolding, Java and framework upgrades, OTel setup, CI setup, production troubleshooting

Stack

Java 25 Kotlin Quarkus Spring Boot Apache Kafka Redis MongoDB AWS EKS AWS Lambda Kubernetes Docker Tencent Cloud GitLab CI OpenTelemetry Grafana GraalVM JetBrains Koog Spring AI OpenSearch Hexagonal Architecture DDD TDD

Impact

Built the platform from zero; it now processes millions of integration events a day across 30+ marketplace channels
Quarkus native migration cut service memory usage by over 60%
Event-driven design holds product, order and stock sync at sub-second latency
Established the engineering culture: DDD, Hexagonal Architecture and automated testing as the default
Turned one-off expert work (upgrades, migrations, new integrations) into repeatable, AI-assisted playbooks the whole team runs

Tech Lead

Serasa Experian Contract · Remote

Concurrent engagement

Jan 2025 – May 2025

Led a biometrics platform team of 6 engineers and 1 QA, working alongside other product teams, at Brazil’s largest credit bureau
The platform verifies identity through facial biometrics and liveness for account opening and fraud prevention, used by clients such as Vivo, Claro and Itaú
Owned the architecture and delivery quality of the team’s microservices; designed asynchronous processing on AWS SQS/SNS
Coordinated technical alignment across teams and with product and business stakeholders

Java Spring Boot AWS SQS/SNS Oracle Autonomous DB Microservices

Context

Serasa Experian is Brazil's largest credit bureau. I joined the Biometrics Product team as Tech Lead, on a five-month contract running alongside my role at GuBee, leading 6 engineers and 1 QA next to other product teams. The platform verifies identity through facial biometrics and liveness checks, for account opening and fraud prevention at clients such as Vivo, Claro and Itaú.

Responsibilities

Lead the team building the biometrics verification platform used by financial institutions and fintechs
Define system architecture and hold the quality bar across the microservices
Run technical alignment and requirements refinement sessions with product and business stakeholders
Turn business requirements into technical work the team could deliver
Design asynchronous messaging flows on AWS SQS/SNS for biometric data ingestion and processing

Stack

Java Spring Boot AWS SQS AWS SNS Oracle Autonomous Database Microservices REST APIs Docker

Impact

Worked on an identity verification platform used by large telecom and banking clients (Vivo, Claro, Itaú)
Kept product, business and the other engineering teams aligned on what we were building and when

System Architect

Atos Londrina / São Paulo

Dec 2014 – Jul 2019

Architect on enterprise integration programmes for NET/Claro, Multiplus Fidelidade and Carrefour
Specialised in Oracle SOA Suite middleware, building OSB, BPEL and BPM interfaces to EAI patterns
Led the product definition for Carrefour's HeavyGoods platform, built on event-based microservices

Oracle SOA Suite OSB BPEL Java Microservices EAI

Context

Atos is a global IT services company. Across nearly five years I worked as a system architect embedded in major enterprise clients: NET/Claro in telecom, Multiplus Fidelidade in loyalty, Carrefour in retail. The role combined architecture definition, team leadership, technical proposals and hands-on integration work on Oracle's enterprise middleware stack.

Responsibilities

Define system architecture and lead development on large-scale enterprise integration projects
Write technical and commercial proposals, and defend the solution in client sessions
Design and implement OSB, BPEL and BPM interfaces following VETO and Enterprise Application Integration practice
Lead cross-functional teams across several client engagements at once
Drive Carrefour's HeavyGoods product definition, improving usability and sales conversion through event-based microservices

Stack

Java Oracle SOA Suite Oracle Service Bus (OSB) BPEL BPM Microservices EAI Patterns REST / SOAP Oracle Database

Impact

Worked as architect for three large clients (NET/Claro, Multiplus, Carrefour), often at the same time
Defined the HeavyGoods product for Carrefour on event-based microservices
This is where I learned the integration patterns I later used to design GuBee

Software Architect & Project Lead

Infracommerce Consultant · Greater Londrina

Nov 2013 – Nov 2014

Led projects on the ACEC e-commerce platform for Carrefour Group, Decathlon and Lança Perfume
Built a service-oriented architecture with separate administration, sales and ERP integration layers
Implemented Drools for dynamic pricing rules and Lucene/Solr for search and product discovery

Java Oracle ADF JSF BPEL OSB Drools

Context

Infracommerce is one of Brazil's leading full-service e-commerce operators. I joined as project leader and software architect for ACEC, a white-label e-commerce platform powering major retail brands, built on a service-oriented architecture with clearly separated layers for administration, sales and ERP/legacy integration.

Responsibilities

Lead development and architecture for the ACEC platform serving Carrefour Group, Decathlon and Lança Perfume
Design the administration layer in Oracle ADF for deal management, customer registration and product maintenance
Optimise the sales layer with REST APIs, JSF and Lucene/Solr for search and product discovery
Implement a Drools rules engine for dynamic pricing and promotion rules
Build the integration layer with BPEL and OSB services exchanging data with client ERPs

Stack

Java Oracle ADF JSF Solr / Lucene Drools BPEL OSB Oracle Database REST SOAP

Impact

The platform ran online stores for large retail brands
With Drools, the business team could change pricing rules without a deploy

Tech Lead

Accurate Londrina / São Paulo

Aug 2010 – Nov 2014

Team lead on CPQD projects delivering city administration systems for the public sector
Owned requirements analysis, function-point estimation, architecture definition and team coordination

Java Architecture Function Points Team Leadership

Context

Accurate was a software consultancy later acquired by Infracommerce. I led development teams on government projects for CPQD, one of Brazil's largest telecom R&D centres, building city administration systems. The role sharpened my requirements engineering, estimation and team leadership.

Responsibilities

Lead the development team on public-sector systems projects
Gather and analyse customer requirements, translating them into technical specifications
Estimate project scope using Function Point Analysis
Define system architecture and technology stack for each project
Coordinate development activities and hold the delivery quality bar

Impact

Delivered multiple city administration systems for CPQD
Built the estimation and requirements-engineering competency that carried through the rest of my career

System Architect

Atos Embedded at Serasa · São Paulo

Jun 2013 – Nov 2013

Architecture definition and technical specification for credit applications, embedded in Serasa's engineering teams

Java Architecture Credit Systems

Context

A short engagement at Atos embedded within Serasa (now Serasa Experian), Brazil's largest credit bureau. My first contact with credit data, twelve years before I went back to Serasa as Tech Lead in 2025.

Responsibilities

Define system architecture for credit-related applications
Produce technical specifications and architecture documentation
Work with Serasa's internal engineering teams on integration points

Tech Lead

LINT Londrina

Jun 2009 – Nov 2010

Led development of a city administration system for CPQD, from requirements to delivery
Owned customer requirements, architecture definition and team coordination through every project phase

Java Team Leadership Architecture

Context

LINT was a Londrina software house where I served as Tech Lead on government projects. The main project was a city administration system for CPQD, which I followed from requirements to delivery.

Responsibilities

Lead the development team through every phase of the CPQD city administration project
Gather and document customer requirements
Estimate project scope using function point analysis
Define system architecture and guide the team through implementation

Systems Analyst

Amadeus IT Group Travel & tourism technology

2008

Built insurance applications and a tourism platform project for Carrefour
Worked across the Java EE stack of the era: Struts 2, EJB3, JPA, Spring, Maven and web services

Java Struts 2 EJB3 JPA Spring

Context

Amadeus IT Group is a global technology company serving the travel and tourism industry. The role deepened my grounding in requirements definition, architecture design and Java enterprise development.

Responsibilities

Develop insurance applications using enterprise Java technologies
Contribute to a tourism platform project for Carrefour
Design and implement REST and SOAP web services for system integration

Stack

Java Struts 2.0 EJB3 JPA Spring Maven Web Services

Team Leader & Quality Assurance

BSI Tecnologia Banking systems · São Paulo

2003 – 2008

Started as a developer and grew into team leader and quality assurance over five years
Delivered projects for Itaú and Bradesco, integrating Java applications with mainframe Cobol systems

Java EJB2 Struts Hibernate Cobol

Context

BSI Tecnologia was a São Paulo software company specialised in financial sector applications. This is where my career began. I went from junior developer to team leadership and quality assurance over five years working with some of Brazil's largest banks.

Responsibilities

Develop and maintain Java applications for Itaú and Bradesco banking systems
Integrate Java applications with mainframe Cobol legacy systems
Lead development teams on banking project deliveries
Establish quality assurance processes and run code reviews

Stack

Java EJB2 Struts Hibernate Cobol Integration Oracle Database Mainframe

Impact

Progressed from junior developer to team leader and QA within the same company
Learned enterprise Java and banking systems here
Integrated Java applications with mainframe COBOL, which not many developers did at the time

Skills

Core

Used daily in production

Java Kotlin Quarkus Spring Boot Apache Kafka Microservices REST APIs DDD Hexagonal Architecture TDD

Applied AI

LLM systems in production

Spring AI JetBrains Koog OpenAI Multi-agent systems RAG pgvector MCP Guardrails DeepEval LLM observability Claude Code

Cloud & delivery

Where it runs and how it ships

AWS Kubernetes Docker AWS EKS AWS Lambda AWS SQS/SNS Tencent Cloud GitLab CI OpenTelemetry Grafana

Data & integration

Storage and enterprise middleware

MongoDB Redis PostgreSQL / PostGIS Redis Streams Oracle Database Oracle SOA Suite OSB / BPEL Solr Drools

Certifications

Sun Certified Enterprise Architect

Sun Microsystems · 2010

Oracle Certified Professional, Java SE 7

Oracle · 2014

Professional Scrum Master I

Scrum.org · 2018

Also certified

Oracle SOA Suite 11g EssentialsOracle

2015

ITIL Foundation CertificateIT service management

2013

Sun Certified Associate, Java PlatformSun Microsystems

2008

Sun Certified Business Component Developer, Java EE 5Sun Microsystems

2007

Sun Certified Business Component Developer, Java EE 1.3Sun Microsystems

2007

Sun Certified Web Component DeveloperSun Microsystems

2006

Sun Certified Programmer, Java PlatformSun Microsystems

2006

Languages

Portuguese Native

English Upper-intermediate — CEFR B2 EF Level 12, EF Education First, 2017

Education

Postgraduate, Computer Software Engineering

Unifil — Centro Universitário Filadélfia

2014 – 2015

Bachelor of Science, Information Systems

Universidade Tecnológica Federal do Paraná (UTFPR)

2003 – 2007

renatojava.github.io · Full profile, project detail and contact links online.

Renato Garcia

About

Experience

Founding Software Architect

Senior Staff Engineer

Tech Lead

System Architect

Software Architect & Project Lead

Tech Lead

System Architect

Tech Lead

Systems Analyst

Team Leader & Quality Assurance

Skills

Core

Applied AI

Cloud & delivery

Data & integration

Certifications

Sun Certified Enterprise Architect

Oracle Certified Professional, Java SE 7

Professional Scrum Master I

Languages

Education

Postgraduate, Computer Software Engineering

Bachelor of Science, Information Systems

Get in touch