I am an AI graduate with passion and hands-on experience in Machine Learning, forecasting, Regression models, RAG systems, Agent Orchestration via LangGraph, and Distributed Systems with Azure Logic Apps. Currently working as an AI/Software Developer at Nedstar in Amsterdam, Netherlands, where I have experience in Business Process Automation.
I am passionate about complex ML models like ST-GNNs and enjoy designing solutions through innovative model architectures. In my free time I also read the newest research papers in the field.
- BSc in Artificial Intelligence - Vrije Universiteit Amsterdam, The Netherlands (2024)
- Focus on: Intelligent Systems, applied machine learning and software applications
- Advanced topics: Computational Intelligence, Machine learning, Text mining (NLP), Software Engineering, Knowledge representation, Logic, Machine Learning, Statistics
- Bachelor's in Management Engineering - Uninettuno University, Italy
- Mathematics, Physics, Business Administration, Statistics, Thermodynamics
- Advanced English Course - TM International School of English, Cambridge, UK
AI/Software Developer at Nedstar (2025) - Amsterdam, North-Holland, Netherlands
- Independently automated invoice processing for 6,500+ documents/year using AI, cutting manual effort by 70%
- Deployed intelligent document processing with custom ML models trained on 2 years of historical data, automating 95% of invoices
- Integrated RAG system with Business Central & Azure, implementing risk-based approval workflows with Blob Storage with consecutive real-time stream of changes via webhook, CosmosDB for state management and FastAPI for serving
Enterprise-scale document processing system automating 6,500+ invoices annually with 95% automation rate
Tech Stack: 
Key Achievements: 70% reduction in manual effort • 95% automation rate • Real-time processing
Impact: Processing 6,500+ documents/year with intelligent risk-based approval workflows
Technical Architecture
RAG Pipeline: Custom ML models trained on 2 years of historical invoice data Integration Layer: Business Central & Azure Blob Storage with real-time webhooks State Management: CosmosDB for persistent workflow state and audit trails API Layer: FastAPI for high-performance document processing endpoints Orchestration: LangGraph agents for intelligent workflow routing and validation
Key Innovations: Risk-based approval routing with ML-driven confidence scoring Real-time stream processing for immediate invoice status updates Custom OCR pipeline optimized for invoice layouts and formats Automated vendor master data reconciliation and validation
Multi-agent system for business process automation with distributed coordination
Tech Stack: 
Features: Multi-agent coordination • Persistent memory • Tool orchestration
Agents: Document Parser • Data Validator • Business Logic • Workflow Controller
Agent Architecture
Coordination Layer: State machines for complex workflow orchestration Memory Management: Redis-backed persistent conversation and context memory Tool Integration: 4 specialized tools for data processing and external system integration Monitoring: Real-time agent performance tracking and decision logging
Agent Specializations: Document Agent: Multi-format parsing, structure extraction, content validation Validation Agent: Business rule enforcement, data quality checks, compliance verification Integration Agent: ERP system connectivity, API orchestration, data synchronization Monitoring Agent: Performance tracking, anomaly detection, alert management
Advanced time series forecasting with hierarchical LSTM for complex pattern recognition
Tech Stack: 
Focus Areas: Hierarchical LSTM • Multi-variate forecasting • Spatial-temporal modeling
Research: Novel architectures for complex pattern recognition in time series data
Research & Innovation
Hierarchical LSTM Models: for capturing complex dependencies on three time levels. Forecasting Pipeline: End-to-end system for multi-horizon prediction tasks Model Architecture: Custom attention mechanisms for temporal and spatial relationships Evaluation Framework: Comprehensive benchmarking against traditional and modern methods
Technical Contributions: Novel graph construction methods for time series relationships Attention-based temporal modeling with memory mechanisms Multi-scale feature extraction for diverse forecasting horizons Production deployment patterns for real-time inference
Enterprise integration platform with Azure Logic Apps for seamless business process automation
Tech Stack: 
Solutions: Workflow automation • System integration • Event-driven architectures
Integrations: ERP systems • CRM platforms • Document management • API orchestration
Integration Architecture
Workflow Engine: Azure Logic Apps for complex business process orchestration Event Processing: Real-time event streaming and processing pipelines API Management: Centralized API gateway with authentication and rate limiting Data Transformation: ETL pipelines for data harmonization across systems
Key Implementations: Multi-system data synchronization with conflict resolution Automated approval workflows with escalation rules Real-time monitoring and alerting for business processes Scalable integration patterns for enterprise applications
