GenAI Engineer @ Datapizza · AI engineering · GeoAI · Sports tech
I build AI applications, data pipelines, and tools for real workflows — from searchable knowledge bases to satellite datasets and volleyball analytics. I also design and deliver hands-on training that helps teams put AI into practice.
My background combines Mathematical and Geoinformatics Engineering at Politecnico di Milano. I work across Python, applied ML, and full-stack development, connecting models and data to interfaces people can use.
Explore NASA near-Earth objects through an orbital view, charts, filters, and detail pages. Next.js · FastAPI · ECharts — source & architecture. The backend may take a little time to wake up on the first visit.
Quaderno: the Claude courses →
An Italian reading experience with chapters, notes, bookmarks, and text-to-speech. An independent adaptation of public Anthropic learning materials. JavaScript · React · GitHub Pages — source.
Articles on AI engineering, multi-agent systems, retrieval, and geospatial data, with a bilingual Markdown publishing pipeline. Node.js · Gemini · GitHub Pages — source.
| Project | What it shows |
|---|---|
| VolleyRef | PDF/OCR extraction → editable, validated volleyball match data → Excel/CSV. Next.js, FastAPI, OpenCV, Tesseract, and an Electron desktop package. |
| datapizzaAI-RAG | A documentation assistant with ingestion, semantic retrieval, query rewriting, memory, and MCP integration. Python, datapizza-ai, Qdrant, Gemini, and Streamlit. |
| Point-cloud research | Thesis work on outliers, denoising, and vegetation removal for illegal-landfill detection, with Python experiments and synthetic data generation. |
| EO dataset pipeline | Sentinel-2 discovery → cloud-aware preprocessing → Zarr/Parquet datasets, quality checks, and a PyTorch data loader. |
| DevPath | An adaptive coding-practice MVP: diagnostic onboarding, progressive hints, and isolated Python execution. Next.js, PostgreSQL/Supabase, and Docker. |
| Blender workflow plugin | AI-assisted 3D production with modelling, UVs, rendering, visual review, and Python checks for geometry, printability, and export. |
My private projects also cover:
- Agent engineering: MCP tools/resources, harness design, skills, deterministic controls, human approval gates, and regression checks.
- Evaluation & platforms: retrieval and generation evaluation, asynchronous scoring, queues, event administration, and role-based access.
- Analytics & reporting: configurable survey pipelines, clustering, LLM-assisted text analysis, traceable artefacts, and automated reports.
- Sports products: touchscreen scouting, rotations and libero logic, video-analysis workflows, and prototypes for automated clip generation.
- Applied ML: 3D segmentation, thermal-image processing, and comparisons between CNN baselines and foundation-model approaches.
- Integrations: OAuth, PostgreSQL/Supabase, Redis/RQ, Google Apps Script, Drive, and calendar workflows.
This work informs my skills; client information, datasets, and private code stay private.
AI & agents
Retrieval, agent orchestration and evaluation · pandas & scikit-learn · OpenAI, Gemini & Anthropic.
Applications & infrastructure
React interfaces · REST APIs & OAuth · Redis/RQ · GitHub Actions.
Geospatial & visual computing
Earth Observation · xarray & Zarr · 3D point clouds · Plotly & Streamlit.
Source: GitHub · Updated daily · Repository metrics cover public work; the language chart counts non-fork repositories.
I enjoy the full path from understanding a problem to building, evaluating, and explaining a solution. Outside work, you'll often find me on a volleyball court.
Browse all public projects · Let's connect
Updated 10 October 2026 · Public examples, plus a capability summary of private work.