Building scalable AI systems and RAG architectures to power intelligent applications. Bridging the gap between cutting-edge research and production. Based in Napoli, Italy.
I am a passionate AI & ML Engineer dedicated to building intelligent systems that tackle real-world complexities. My expertise lies in bridging the gap between intricate data and actionable software solutions, designing scalable ML pipelines and robust models that drive automation and efficiency.
Currently, I am deeply immersed in the world of Large Language Models (LLMs) and Generative AI, exploring how to make machines understand and interact with us more naturally. From Fine-tuningmodels to implementing RAG architectures, I leverage the latest open-source advancements to integrate smarter, context-aware AI into everyday applications.
Architected and deployed production-ready Hybrid RAG systems, implementing advanced Adaptive RAG architectures to optimize retrieval precision and response relevance.
Designed and integrated scalable microservices for AI applications, ensuring high availability and seamless communication across distributed system infrastructures.
Optimized data storage and retrieval by managing Vector databases alongside structured SQL systems, specifically handling complex nested and layered data strings.
Streamlined the MLOps lifecycle by containerizing AI services with Docker and maintaining rigorous Git version control for end-to-end model deployment.
Leveraged large language models (LLMs) and specialized libraries such as PandasAI to build automated data analysis tools for production-level environments.
Collaborated within cross-functional teams to translate complex AI research into scalable engineering solutions, ensuring alignment with project goals.
Engineered advanced jailbreak techniques to stress-test state-of-the-art LLMs against toxicity and safety guardrails.
Conducted extensive adversarial prompting experiments, enhancing the team's understanding of model bias and failure modes by 53%.
Authored key technical reports on findings and mitigation strategies, serving as a foundation for ongoing Responsible AI standards.
Engineered scalable database architecture using PHP and Laravel, achieving a 65% boost in query performance through advanced SQL optimization.
Developed and implemented robust API integrations, facilitating seamless and secure data synchronization between backend services and frontend interfaces.
Enhanced system stability and reliability, resulting in a 20% reduction in crash incidents during critical stress testing phases.
Specializing in the intersection of Mobile Development and Artificial Intelligence, leveraging Swift and SwiftUI to deploy complex Machine Learningmodels on iOS devices.
Developed an intelligent financial assistant app utilizing the Vision Framework for OCRand a custom ML classifier to scan receipts, categorize expenses automatically, and visualize spending habits via dynamic charts.
Engineered a pet-wellness application powered by Deep Learning and Computer Vision, capable of analyzing canine facial expressions to detect and interpret emotions in real-time.
Focusing on On-device Machine Learning (Edge AI) to ensure user privacy and high-performance inference without server dependency.
Thesis: 'Causal Toxicity Testing of Large Language Models'.
Grade:100/110.
Focus:AI Safety, Large Language Models (LLMs), and Causal Inference.
Research:Conducted in-depth research on LLM vulnerabilities, developing methodologies to identify and mitigate toxicity generation through causal testing frameworks. This work directly aligns with Red Teaming and AI Alignment practices.
Relevant Coursework: Machine Learning, Deep Learning, Natural Language Processing (NLP), Statistical Learning, Big Data Analytics.
Capstone Project: Content Management Systems utilizing PHP.
Grade:16.34/20.
Capstone Project: Designed and implemented a custom Content Management System (CMS) using PHP and MySQL. Focused on architecture patterns and database efficiency.
Core Competencies: Data Structures & Algorithms, Object-Oriented Programming (OOP), Software Architecture, Database Design.
Generative AI | Multi-Agent RAG | OpenAI & Llama | FAISS | RAGAS Evaluation
Engineered a scalable Multi-Agent RAG architecture for civil law, employing a supervisor-worker pattern to coordinate localized retrieval across multiple jurisdictions.
Achieved superior grounding with 0.867 Context Recall and 0.776 Faithfulness scores by implementing hybrid routing strategies and validating performance via the RAGAS framework.
View ProjectMachine Learning | Scikit-Learn | Data Imbalance Handling | Medical Data Analysisp>
Spearheaded the development of a statistical machine learning model to predict ALS King’s clinical stages, utilizing Ordinal Logistic Regression to accurately capture the hierarchical nature of disease progression.
Engineered a robust data pipeline using Python (Pandas, Scikit-learn), implementing SMOTE to resolve severe class imbalance and improve prediction accuracy for underrepresented patient groups.
View ProjectDeep Learning | TensorFlow | XGBoost | PyTorch | Hyperparameter Tuning
Developed a high-performance binary classification system to forecast rainfall, processing over 145,000 weather records and handling significant class imbalance.
Engineered an Artificial Neural Network (ANN) with entity embeddings using TensorFlow/Keras, achieving a best-in-class F1-Score of 0.66 and 84.7% Accuracy, outperforming Random Forest and Logistic Regression baselines.
View ProjectStreamlit | Data Visualization | Unsupervised Learning | Web Application Development | PCA
Developed an interactive machine learning web application using Streamlit to visualize and predict student academic outcomes, translating complex educational data into actionable insights.
Implemented unsupervised learning techniques including K-Means Clustering and PCA to segment students into distinct behavioral groups for targeted academic interventions.
View ProjectRegression Analysi | Ensemble Methods | Feature Engineering | Predictive Analytics | XGBoost
Engineered a high-precision regression model to forecast used car prices, achieving an outstanding R² score of 0.96 and minimizing Mean Absolute Error (MAE) to 0.62.
Developed a Stacking Ensemble architecture combining XGBoost, Random Forest, and Linear Regression, outperforming individual models by capturing both linear and non-linear data patterns.
View Project