AI/ML Engineer building mission-critical GenAI systems in regulated fintech environments.
- 🛡️ Designing LLM security middleware — adversarial defense, PII anonymization, prompt lifecycle auditing
- 📉 Cut 85% inference costs by replacing managed services with in-house distilled models
- 🚀 Built RTGen (INPI-registered software) — GenAI report generation 39x faster than human baseline
- 🎓 M.Sc. Computer Science + B.Sc. Statistics — Federal University of Amazonas
|
Banco Cora — Fintech End-to-end security layer for LLM interactions in a regulated banking environment.
95% attack detection | 96% PII accuracy | 14 sensitive data types |
Registered Software (INPI) GenAI-powered technical report generation system using RAG architecture.
39x faster than human | 15% OpEx savings |
|
1st Place — Brazilian ML Olympiad 2024 Optimized BERT model that outperformed commercial LLM-based solutions for toxic language detection in PT-BR.
1st place | Cost-efficient | Beat commercial LLMs |
Instituto de Pesquisas Eldorado Re-engineered NLP inference pipelines replacing expensive managed services with distilled in-house models.
85% cost reduction | 92%+ accuracy maintained |
Models: Gemini GPT LLaMA Claude BERT GLiNER Grok
Techniques: Fine-tuning (LoRA/QLoRA) RAG Red Teaming Purple Teaming LLM-as-a-Judge
Methods: Bayesian Inference A/B Testing SHAP Classification Regression Hybrid Filtering
AI Engineer & Solutions Architect 2025 - Present
Security middleware for Generative AI in a highly regulated financial environment
- Designed async processing core for message interception with LLM provider decoupling
- Led Red Teaming strategies and fine-tuning for adversarial attack detection (Jailbreaks/Prompt Injection)
- Built proprietary NER models for real-time anonymization of 14 sensitive data types (LGPD compliance)
- Established Tracing/Logging standards for full prompt lifecycle auditability
Senior Software Analyst 2021 - 2025
AI solutions for global tech players focused on ROI maximization
- Re-engineered NLP pipelines: 85% cost reduction replacing managed services with distilled models (92%+ accuracy)
- Created RTGen (INPI-registered): GenAI report system 39x faster than human, 15% OpEx savings
- Optimized recommendation algorithms (Hybrid Filtering): 3% CTR uplift on end-user engagement
- Structured scalable ETL/ELT pipelines and CI/CD for ML with scientific reproducibility
| Achievement | Year | |
|---|---|---|
| 🥇 | 1st Place — Brazilian Machine Learning Olympiad — Toxic language detection, outperforming commercial LLMs with optimized BERT | 2024 |
| 🏆 | EldFlash Award — E2E architecture with LLMs for strategic risk metric | 2025 |
| 🌟 | Eldorado Excellence Award — LLM automation architecture with high ROI | 2024 |
| 📄 | Registered Software (INPI) — Author of RTGen, AI-powered report generation | 2024 |
| ⚡ | EldFlash Award — Tracking and resolution of critical pipeline failure | 2022 |
| 🎓 | M.Sc. Computer Science — Federal University of Amazonas | 2021 |
| 📊 | B.Sc. Statistics — Federal University of Amazonas | 2019 |
