Brazil
Information Systems student at UNIRIO passionate about transforming complex data into intelligent solutions. With a strong focus on Data Science, Data Engineering, and Artificial Intelligence, I aim to build scalable architectures and efficient predictive models. I have hands-on experience developing robust data pipelines, RAG (Retrieval-Augmented Generation) systems, and integrating Multi-LLM ecosystems—including the deployment of cutting-edge open-weights models like DeepSeek and Qwen. My core tech stack includes Python, SQL, PyTorch, Keras, and TensorFlow, combined with practical experience in infrastructure and cloud computing. Driven by technical challenges, I am actively seeking internship opportunities (Data Engineering, Data Science, or Machine Learning) where I can design end-to-end solutions and deliver real-world business impact.
Researcher on the AutoNEM project, focusing on Natural Language Processing (NLP) for extracting and structuring events in journalistic texts using the 5W1H framework. • Developed and advanced the architectural design of Deep Learning models utilizing PyTorch and Keras. • Achieved significant reductions in training and inference times by applying advanced model compression techniques and hyperparameter fine-tuning. • Trained, generated, and semantically validated domain-specific embeddings, conducting comparative analyses against industry pre-trained models (e.g., Google News) to optimize vector representativeness. • Optimized data processing scripts and maintained strict version control to ensure the reproducibility of scientific experiments.
Planned and taught chess lessons
Direct involvement in the data ecosystem, focusing on building, modernizing, and maintaining ingestion pipelines. • Developed and optimized data ingestion workflows into Data Lakes utilizing Apache Hive and AWS cloud infrastructure. • Increased agility and efficiency in creating new data pipelines, ensuring data quality, governance, and high availability. • Actively troubleshooted bottlenecks and mitigated structural issues, ensuring robust database reliability for internal clients.
End-to-end design and development of a "Smart Document Analyzer", an AI-driven full-stack web application for querying PDF databases. • AI Engineering & RAG: Implemented a comprehensive RAG pipeline using LangChain and FAISS for vectorization and high-precision semantic search. • Multi-LLM Orchestration: Built a modular architecture dynamically integrating Google Gemini, OpenAI GPT, and Anthropic Claude, featuring an automated API key rotation system to bypass rate limiting. • Search Optimization: Engineered an LLM-powered query expansion mechanism, significantly improving the relevance of retrieved contexts. • Infrastructure & Deployment: Structured a Flask backend and a responsive HTML/JS frontend (featuring real-time response streaming), securely deploying the application publicly via Google Colab and Cloudflare Tunnel.