AI/ML Engineer | ML Platform Engineering | MLOps | LLMOps | AI Infrastructure | SRE
📍 Garanhuns, Pernambuco, Brazil
💼 AI/ML Engineer @ CashMe | Cyrela Group
🎓 B.S. in Software Engineering @ UPE
🎓 Enrolled in a Postgraduate Specialization in Artificial Intelligence Engineering & MLOps @ PUC Minas · Starting Sep 2026
📫 thiago.barros.empr@gmail.com
🔗 LinkedIn · Medium
I build the platforms, infrastructure, and reliability foundations that turn Machine Learning models and AI agents into secure, observable, and scalable production systems.
Currently, I work as an AI/ML Engineer at CashMe, part of Cyrela Group, across Machine Learning Engineering, ML Platform Engineering, MLOps, LLMOps, AI Infrastructure, and Site Reliability Engineering.
My work covers the full production lifecycle of AI systems: Infrastructure as Code, CI/CD and GitOps, model and agent serving, automated evaluation, observability, security, governance, FinOps, monitoring, retraining, and developer experience.
I care about building AI systems that engineering teams can trust, operate, and continuously evolve, not one-off demos.
- Automating credit analysis and underwriting workflows using Machine Learning, credit scoring, income estimation, payment behavior, collateral, and other risk signals.
- Building a centralized MLflow-based observability and evaluation platform for AI agents and ML models.
- Implementing golden-dataset evaluations, LLM-as-a-Judge, regression testing, A/B experiments, tracing, and automated agent and prompt versioning.
- Developing CI/CD foundations for a monorepo-based agent ecosystem, including unit and integration tests, automated dependency updates, dynamic agent discovery, and reusable tools.
- Reduced new AI solution setup time from 4–5 months to one month or less through reusable OpenTofu and Terraform modules, standardized FastAPI templates, and production engineering standards.
- Reduced AWS infrastructure costs by 55.56% in a regulated fintech environment while preserving the reliability, performance, and security of critical systems.
- Architected and implemented an AI platform featured in an official Google Cloud customer success case, achieving 93% classification accuracy and processing more than 3,000 images per execution.
- Built a secure MCP-based NL2SQL solution that reduced business time-to-insight from 4–6 hours to approximately 30–90 seconds.
- Developed reusable GenAI engineering standards and SDKs that increased developer productivity by 35% and reduced incidents in a shared Google Cloud environment by more than 80%.
| Project | Description |
|---|---|
| Google Cloud Customer Success Case | Production multimodal AI platform using Gemini, FastAPI, and Cloud Run Jobs, achieving 93% classification accuracy and processing thousands of images per execution. |
| PDF Preprocessing Pipeline | OCR and dataset-generation pipeline for document classification and information extraction. |
| Event-Driven Microservices | Java microservices architecture using Spring Boot, RabbitMQ, Docker, Kubernetes, PostgreSQL, and GitHub Actions. |
| Customer Churn Analysis | Telecom churn analysis recognized as the top initiative in the PET Data Analytics and Machine Learning program. |
- B.S. in Software Engineering, University of Pernambuco
- Enrolled in a Postgraduate Specialization in Artificial Intelligence Engineering and MLOps, PUC Minas · Starting Sep 2026 · Expected Nov 2027
- Google Cloud Professional Data Engineer
- GDG Garanhuns Organizer
- 1st Place, PET Data Analytics and Machine Learning Program
I write about AI/ML Engineering, MLOps, LLMOps, Software Engineering, Cloud, and the challenges of operating reliable AI systems in production.
Thanks for stopping by :D!