Resume

PhD candidate in Computer Science (Machine Learning Engineering)

Rio de Janeiro, Brazil

MBA in Data Science & Analytics

Piracicaba, Brazil

M.Sc. in Physics (Particle Physics)

Rio de Janeiro, Brazil

B.Sc. in Physics with Teaching Certification

Niterói, Brazil

Applied Scientist & AI Engineer

Ottawa, Canada

• Conduct applied research on multimodal generative model inference, scalability, and AI systems performance. Proposed adaptive routing to select image-processing configurations based on task context, and developed semi-analytical models for latency, resource utilization, and serving capacity.
• Built a capacity-planning platform for AI datacenters to identify hardware configurations that meet SLOs for multimodal serving; characterized inference bottlenecks and developed strategies to improve latency, scalability, and hardware efficiency.
• Technologies: Python, FastAPI, React, Hugging Face, vLLM, PyTorch, Docker, NVIDIA Nsight Systems, and NVIDIA AI Perf.

Machine Learning Engineering Doctoral Researcher

Rio de Janeiro, Brazil

• Investigate agile management and requirements engineering practices for ML workflows, translating engineering challenges into evidence through industry–academia collaboration.
• Delivered MVPs for industry partners; received the Best Industrial Experience Paper Award at Software Quality Days 2025 for work on ML process optimization.

Data Scientist

Rio de Janeiro, Brazil

• Shipped an agentic cybersecurity service connecting generative AI with proprietary fraud, malicious-site, and financial-transaction tools using LangChain, PyTorch, YOLO, Scikit-learn, RAG, and Model Context Protocol (MCP). Replaced third-party services with in-house systems and applied Clean Architecture with FastAPI and NestJS.
• The service reached approximately 4 million users; its fraud-detection pipeline outperformed the legacy solution by more than 60%, while eliminating third-party operating costs.
• Modernized an NLP and supervised-learning service, achieving approximately 79% on its evaluation metric (+14%) while reducing the model footprint by tens of MB.
• Built a computer-vision system using fine-tuned YOLO models to interpret electrical diagrams, reducing repetitive manual work by approximately 10 hours per week.
• Led delivery from discovery to production as Agile lead and Solutions Architect. Built production systems on GCP with Docker and conducted data analysis with Pandas and SQL.
• Technologies: Python, PyTorch, Scikit-learn, YOLO, LangChain, MCP, SQL, Docker, GCP, FastAPI, NestJS, and TypeScript.

Data Scientist

Geneva, Switzerland

• Designed and validated ML analytical pipelines using statistical inference, hypothesis testing, and uncertainty analysis. Ran more than 4,000 Monte Carlo pseudo-experiments to assess robustness, bias, and signal-to-noise separation.
• Validated an unbiased strategy with zero spurious signals, reduced pipeline memory by 80% (approximately 100 GB to 20 GB RAM), and saved more than 2 TB of storage.
• Technologies: Python, C++, ROOT, Pandas, NumPy, Scikit-learn, Linux, Bash, and Monte Carlo simulation.