ProfilePhysics-trained ML Engineer with 5+ years across industry and consulting. Builds custom deep learning systems, reinforcementlearning agents, and complex-systems models, with the production engineering (cloud, edge, HPC) to ship them. BSc and MSc in Physics from Padova, both with theses on neural network training on enterprise NVIDIA GPUs. Has delivered production ML for global manufacturing, academic research, physics and finance, with full-stack experience on Azure, GCP and Databricks on AWS and 9+ years of advanced Python.
Machine Learning Engineering- Physics-Informed AI
- Differential Equations, PINNs, Neural ODEs, Bayesian Networks/Statistics, ML/DL
- Complex Systems Modeling
- Dynamical Systems, Network Analysis, Agent-Based Models, State Reconstruction
- Agentic Systems
- Multi-step orchestration, tool use, structured outputs, evaluation pipelines
- Reinforcement Learning
- Q-Learning, PPO optimization, Gym APIs, Reward Shaping
- High-Performance ML
- CUDA, MPI, FPGA Data Acquisition Systems
- Generative AI
- RAG Architectures, Azure Cognitive Services, LangChain agents
- Time Series Analysis
- ?TFT, N-BEATS, Prophet, Hierarchical Modeling
Cloud & MLOps- Orchestration
- Kubernetes, Docker, Kubeflow
- GPU Computing
- Hands-on experience with enterprise NVIDIA GPUs, distributed training
- Cloud AI Services
- Databricks on AWS, AWS SageMaker training, deployment, and MLOps patterns, Azure Functions, Azure AI Search, GCP Vertex AI, Databricks, others
- Infrastructure as Code
- Terraform, Ansible, ARM Templates
- Data Pipelines
- Spark, Dask, Ray, Airflow
- Linux Environments
- 10+ years as primary OS (Debian/Arch), Raspberry Pi, HomeLab
- Enterprise Integration
- ?SAP ERP, REST APIs, SQL data pipelines
Tools- Scientific Libraries
- NumPy
- SciPy
- PyTorch
- Scikit-learn
- Statsmodels
- many others
- IoT & Protocols
- MQTT
- ESP32/Embedded C
- REST APIs
- DevOps & Utilities
Domain KnowledgePhysics background applicable across scientific and industrial domains. Hands-on experience in:
- Complex Dynamical Systems Modeling
- Geophysical, Environmental, and Signal-Processing Systems
- Industrial-Scale Data Architectures
- IoT and Predictive Maintenance Solutions
- Enterprise integration (SAP ERP, REST APIs, SQL workflows)
Comfortable ramping on adjacent physics domains (electromagnetism, statistical mechanics, optics, fluid dynamics, materials) in 1-2 weeks.
Core Expertise- Physics-Informed Modeling: Develop predictive systems for complex dynamical systems, including state reconstruction from sparse data, network and agent-based models, and physics-informed neural architectures. Bayesian inference, differential equation-based modeling, transformers.
- Cloud-Native MLOps:
- Designed GPU-accelerated pipelines on bare-metal Kubernetes clusters (E4/CERN infrastructure)
- Deployed scalable ML workflows on Azure/GCP using Databricks, serverless components, and large-scale data architectures
- Industrial AI Solutions: Delivered RAG and document AI at multi-terabyte scale for a Fortune 500 Italian cable manufacturer; transformer architectures and time-series forecasting (TFT, N-BEATS) for seismic signal processing and demand forecasting on distributed HPC.
- Performance Optimization: Applied computational physics methods and HPC deployment experience to enhance model efficiency.
Academic Projects (MSc Physics of Data)- Seismic Pattern Analysis
- Analyzed 30-year earthquake catalog using wavelet transforms and PCA
- Identified scaling laws in waiting time distributions with magnitude-dependent cutoffs
- Tools: NumPy | Report
- Causal Network Inference
- Implemented K2 algorithm with MCMC sampling for Bayesian networks reconstruction (R)
- Compared mutual information vs. factor analysis strategies
- Methods: Markov Chains, Graph Analysis | Slides
- Multi-Agent RL Framework
- Designed reward decay strategies for resource competition
- Built custom Gym environment with experience replay buffer
- Methods: RL, Q-Learning | GitHub
- Epidemic Network Modeling
- Simulated SIS/SIR models on complex networks using Monte Carlo methods
- Analyzed infection thresholds and network robustness to node removal
- Tools: NetworkX, Monte Carlo, Report
- Manifold Visualization
- Designed forced projection technique for color-mapping continuous parameters in latent spaces
- Applied t-SNE/UMAP to complex data for 2D/3D interpretability
- Tools: scikit-learn, Matplotlib, Plotly, GitHub
- Forest Ecosystem Modeling
- Implemented maximum entropy models with Lagrange multipliers for spatial distribution patterns
- Analyzed 100 ha rainforest data through 50 × 50 m2 subplot sampling
- Tools: SciPy, Pandas, GitHub