B.Sc. Artificial Intelligence | Università degli Studi di Pavia
📍 Pavia, Italy

Full video: Open player.

I am an Artificial Intelligence graduate from the University of Pavia with four research internships spanning diverse domains from temporal graph neural networks for financial prediction to multimodal deep learning for cancer immunotherapy. My research journey reflects a deliberate exploration toward understanding principled machine learning methods that bridge theoretical foundations with real-world impact.

Currently focused on computational biology and healthcare AI, I'm particularly interested in deep generative models, multimodal data integration, and uncertainty quantification in high-dimensional biological systems.

I also develop synthetic datasets and evaluation benchmarks for healthcare, scientific, and language applications.

"Each project was a step toward understanding not just how to apply machine learning, but what fundamental principles govern effective learning from complex data."

I develop synthetic datasets and evaluation benchmarks across healthcare, science, and language applications.

• Medical imaging: Retinal fundus and long-bone fracture images.
• Patient trajectories: Oncology, treatment response, survival, and longitudinal measurements.
• Clinical text: Triage records, microbiology reports, antimicrobial resistance, and adverse-event narratives.
• Physiological data: Simulated neural signals and gait telemetry.
• Molecular data: Enzyme sequences, reaction selectivity, and biological composition graphs.
• Medical documents: Fictional credentialing records and evidence reconciliation.

• Language & NLP: Artificial languages, fictional narratives, financial text, and question answering.
• Scientific data: Molecular graphs, chemistry, physical simulations, and mathematical reasoning.
• AI evaluation: Document consistency, retrieval, citation grounding, hallucination, and prompt-injection detection.
• Code & decision tasks: Code transformations, operations, recommendations, and scenario–response pairs.

Formats: Images, text, time series, tables, graphs, and linked documents.

Sample datasets can be provided upon request.

flowchart TD
    subgraph Theory[" 🧮 THEORETICAL FOUNDATIONS "]
        DM[Diffusion Models<br/>Statistical Physics & ML<br/>Prof. Gherardi - UniMi]
    end
    
    subgraph Methods[" ⚙️ METHOD DEVELOPMENT "]
        TGCN[Temporal Graph Networks<br/>Financial Prediction<br/>Prof. Zignani - UniMi]
        CGM[CGM Validation<br/>Pediatric Obesity Study<br/>Prof. Aiello - UniPV]
    end
    
    subgraph Application[" 🎯 CLINICAL APPLICATION "]
        MMIO[Multimodal Immunotherapy<br/>NSCLC Prediction<br/>Dr. Miskovic - PoliMi]
    end
    
    subgraph Synthesis[" 💡 RESEARCH SYNTHESIS "]
        PHD[ Research Vision<br/>Principled ML for<br/>Computational Sciences ]
    end

    DM -->|Training dynamics| MMIO
    TGCN -->|Structured data| MMIO
    CGM -->|Biomedical knowledge| MMIO
    
    DM -->|Theory| PHD
    TGCN -->|Graphs| PHD
    CGM -->|Clinical| PHD
    MMIO -->|Multimodal| PHD

• 🎤 "A Vision-Language Foundation Model for Precision Oncology" (Nature, 2025) - AI-ON-LAB weekly meeting
• 🎤 "Heat Death of Generative Models in Closed-Loop Learning" - Gherardi Group seminar
• 🎤 "Predicting Changes in Glycemic Control from Wearable Device Data" - Medical Applications course

• 📄 CGM vs OGTT Comparative Analysis - Expected submission Spring 2026
• 📄 Bachelor's Thesis: Glucose Response Patterns Based on Macronutrient Composition

🧠 Deep Learning: CNNs, Transformers, VAEs, Diffusion Models, Graph Neural Networks
📊 Statistical Methods: Bayesian Inference, Survival Analysis, Clinical Validation Metrics
🔬 Domains: Computational Biology, Healthcare AI, Quantitative Finance
🧪 Dataset Engineering: Synthetic Data, Dataset Design, Model Evaluation

flowchart LR
    subgraph CB[" 🧬 Computational Biology "]
        CB1[Single-cell Analysis]
        CB2[Spatial Omics]
        CB3[Perturbation Prediction]
    end
    
    subgraph DGM[" 🎲 Deep Generative Models "]
        DGM1[Diffusion Models]
        DGM2[VAEs]
        DGM3[Flow Matching]
    end
    
    subgraph ML[" 🔀 Multimodal Learning "]
        ML1[Cross-modal Fusion]
        ML2[Uncertainty Propagation]
        ML3[Modality Weighting]
    end
    
    subgraph PML[" 📐 Probabilistic ML "]
        PML1[Causal Discovery]
        PML2[Uncertainty Quantification]
        PML3[Bayesian Methods]
    end
    
    CENTER((Research<br/>Vision))
    
    CB --- CENTER
    DGM --- CENTER
    ML --- CENTER
    PML --- CENTER

Core Questions I'm Pursuing:

• How should we weight different modalities when they provide conflicting signals?
• What inductive biases are appropriate for cross-modal representation learning?
• How do we rigorously quantify uncertainty when integrating heterogeneous biological data?