Applied AI Researcher addressing the fundamental tension between interpretability and accuracy in Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) architectures. Specializing in the design of domain-specific AI systems that mitigate hallucination and algorithmic bias in high-stakes educational environments.

Methodological expertise includes constructing vector-based knowledge bases (FAISS), implementing fairness-aware evaluation metrics, and transforming black-box model predictions into transparent, human-verifiable insights for institutional policy decision-making.

• Trustworthy & Responsible AI: Algorithmic fairness, bias mitigation, and safety in decision-support systems.
• Generative AI: Retrieval-Augmented Generation (RAG), LLM fine-tuning, and hallucination reduction.
• Explainable AI (XAI): Interpretability–accuracy trade-offs and model transparency.
• AI in Education: Educational Data Mining (EDM) and automated program evaluation.

• AI & ML Frameworks: LangChain, LlamaIndex, Scikit-Learn, TensorFlow (Basics).
• Vector Databases & Search: FAISS, ChromaDB, Embeddings generation.
• Data Science & Analytics: Python (Pandas, NumPy), R (ggplot2), SQL, Power BI.
• Cloud & Tools: Docker, Google Cloud Platform (GCP), Git/GitHub, Jupyter.

Reconciling Interpretability and Accuracy in RAG Systems for Educational Program Evaluation.

• Institution: Federal University of Rio Grande (FURG), Brazil.
• Research Focus: Investigating the interplay between vector retrieval density and generation quality.
• Objective: Developing a framework to reduce hallucinations in automated academic assessments using CAPES datasets.

🔹 Educational RAG Evaluation System

A RAG architecture using LangChain and FAISS to evaluate Brazilian graduate programs, focusing on hallucination reduction and retrieval accuracy.

🔹 Insurance RAG Prototype

A case study on embedding alignment for legally compliant retrieval in high-stakes financial claims processing.

🔹 Public Health Data Visualization

Comparative visualizations of global pandemic trends using R (ggplot2) to communicate complex health data to non-technical stakeholders.

📄 Download CV | 📧 Email | 🔗 LinkedIn | 📄 Google Scholar

"Transforming black box predictions into transparent, human-verifiable insights."