Senior AI Engineer · Data Scientist
Available for B2B Consulting, Contract Roles, and Freelance Projects.
I partner with US and European teams to architect Knowledge Graphs, AI agents, and production RAG systems that actually ship — currently as an independent contractor for PwC, building enterprise-scale graph solutions for the US market.
Trusted by industry leaders
0+
Years in production AI
0+
Projects shipped to production
0x
Faster knowledge retrieval
0
Cloud & graph certifications
02 / Tech Stack
Opinionated tooling across Python, PyTorch, and Neo4j — optimised for fast iteration, observability, and safe deployment on Azure & AWS.
Core Stack
Built around a lean, battle-tested toolchain for applied ML and agents.
Python
Core language for orchestration, data work, and legacy migrations.
PyTorch
Fine-tuning, custom heads, and experimentation when off-the-shelf falls short.
Neo4j GDS Certified
Graph-native storage for GraphRAG and complex reasoning over industrial processes.
Focus
Where I spend most of my time.
Retrieval-Augmented Generation
Document ingestion, hybrid search, evaluation, and guardrails for reliable answers.
AI Agents
LangGraph-based multi-agent systems orchestrating tools, workflows, and memory.
MLOps
CI/CD, monitoring, and rollback-safe deployments on Azure & AWS (certified).
Stats
Years of experience
4+
Projects delivered
15+
Ecosystem
A tight ecosystem to keep latency low, costs predictable, and iteration loops short.
LLM Providers
OpenAI, Azure AI
Vector / Search
Hybrid search, GraphRAG
Infra
Azure, AWS, Docker, K8s
Monitoring
Metrics, tracing, evals
03 / Process
A streamlined process focused on clear communication, quality delivery, and long-term value.
We start with a call to understand your business goals, technical constraints, and timeline. I ask questions, listen carefully, and define a clear scope.
I design the architecture, build iteratively, and keep you in the loop with regular updates. Clean code, tested logic, documented choices.
We deploy to production with proper CI/CD, monitoring, and documentation. I ensure a smooth handover and remain available for support.
04 / Experience
A quick snapshot of where I've applied AI in production — across industrial automation, retail, consultancy, and education.
Senior Data Scientist & AI Engineer
May 2026 – Present
PwC (Independent Contractor)
United States · Remote
B2B consulting for enterprise-level AI projects targeting the US market. Architecting and deploying highly scalable Knowledge Graphs and AI solutions. R&D on agentic AI frameworks and graph database integrations.
AI Engineer & Data Scientist
Sep 2025 – Apr 2026
Fortil Group (Consultant for Coesia)
Remote / Bologna, Italy
Designed GraphRAG architectures and predictive maintenance pipelines for industrial automation.
Data Scientist
Mar 2024 – Sep 2025
JAKALA
Milan, Italy
Led SAS → Python migration and delivered churn & customer value models on large-scale cloud data stacks.
Data Analytics Instructor
Jan 2024 – Present
Boolean
Milan, Italy (Remote)
Teaching Python, SQL, and ML to 30+ students per cohort with 95% satisfaction, mentoring on best practices.
Data Scientist
Sep 2023 – Mar 2024
Grandi Stazioni Retail
Milan, Italy
Owned end-to-end data pipelines and BI dashboards to monitor retail KPIs across major stations.
Data Scientist & Computer Vision Engineer
Jul 2022 – Sep 2023
Avvale
Milan, Italy
Deployed AWS-based computer vision for textile defect detection and NLP models on customer tickets.
MSc in Physics (Applied Physics)
110/110University of Milan
Specialized in applied physics with a strong focus on Machine Learning and AI techniques for data analysis and modeling.
05 / Blog
TypeSafe AI just launched JEV, a non-generative 'System One' model that is 200x faster and 450x cheaper than traditional LLMs. It recently helped an AI agent beat Pokémon Red. Here is the architectural deep dive on why structured decision models are the missing piece for production-grade autonomous agents and machine-to-machine pipelines.
Read articleOpenAI o1 dropped on September 12, trading intuitive forward passes for hidden chain-of-thought search. But in production, test-time compute introduces a 15–45s latency cliff, 4,000-token hidden billing overheads, and severe inverse-scaling failure modes. Here is the architectural deep dive, the benchmark truth, and the production-grade Cognitive Routing Gateway I deploy to harness System-2 reasoning without blowing enterprise budgets.
Read article06 / Case Studies
Engagements where I shipped agentic systems, GraphRAG, and ML pipelines all the way to production — with measurable outcomes.
80% automation rate
Freed 20+ hours/week of engineering time
Built autonomous AI agents using LangGraph to automate repetitive engineering workflows. Multi-agent system handles document processing, data extraction, and report generation.
Engineering workflows · Automation
Deployed an AWS-based deep learning system for automated textile defect detection. Built with TensorFlow/Keras, processing real-time video feeds for quality assurance and replacing manual inspection.
Textile · Quality control
Automated visual inspection process
07 / Engagement
Clear, outcome-driven engagement models — pick the one that fits where you are, and we scope the rest on a discovery call.
Best for: Teams deciding what to build
A focused review of your data, models, and infrastructure to find the highest-ROI opportunities and de-risk the build before you commit budget.
Best for: Shipping a real system
End-to-end design and delivery of GraphRAG, agentic systems, or ML pipelines — built to production standards, not demos.
Best for: Ongoing momentum
Embedded senior AI engineering for teams that need consistent delivery — I own outcomes, raise the bar, and mentor your engineers.
08 / FAQ
The things teams usually want to know before reaching out. Anything else? Just ask my AI or send a message.
Production AI and data work: GraphRAG and retrieval systems, autonomous AI agents (LangGraph), MLOps pipelines, LLM integration and fine-tuning, computer vision, and data engineering. I focus on systems that ship to production with monitoring and evaluation — not throwaway prototypes.