AI and automation systems builder. I turn messy operational problems into reviewable workflows, internal tools, and focused products.
My background is in growth, copy, and Italian localization for SaaS and fintech. Since 2020 I have increasingly worked in JavaScript, Python, workflow automation, and agent-assisted delivery—using that mix to connect user needs, business processes, and implementation.
| Project | Problem | What I shipped |
|---|---|---|
| AlidTours platform · live preview | Give a small travel business a safer way to manage content and recurring operations. | A Next.js content platform, Git-based editorial review, and four sanitized n8n workflows for lead intake, CRM follow-up, reporting, and shared error handling. |
| Kinetic Monitor Pre-vis | Make an expensive physical-installation concept reviewable before production. | A constraint-driven Python motion lab, diagnostic renders, tests, and a bounded agent-assisted critique loop for a 5×3 kinetic display. |
| Cenere | Turn a discretionary futures method into a deterministic, inspectable research pipeline. | A sandboxed ES/NQ system spanning data ingestion, signal logic, backtesting, risk controls, and a kill-switch, built with a multi-agent development workflow. It is research software, not a live-performance claim. |
| Aerodinamica guidata · live site | Help an aeronautics student move from memorizing formulas to solving problems. | 9 lessons, 40 structured quizzes, 31 worked exercises, 22 generated diagrams, content checks, and automated GitHub Pages deployment. |
Smaller product experiments include Cucina Felice, an offline-first meal planner with ingredient normalization, and a goal-based CrossFit tracker with local-first storage and optional self-hosted sync.
- Start with the process. Find where time is lost, which exceptions matter, and what outcome would count as useful.
- Make boundaries explicit. Separate automation from decisions that still need human review; define privacy, failure, and operating limits early.
- Use the lightest tool that holds up. That may be a script, workflow, interface, agent, or a combination—not a predetermined stack.
- Ship something inspectable. Tests, diagnostics, audit trails, preview environments, and honest limitations matter more than a polished demo alone.
- Measure adoption and impact. A system creates value only when people can use it reliably and the before/after change is visible.
AI is part of my engineering workflow, not a substitute for ownership. I define the problem and constraints, direct the work, review the output, test important paths, and decide what ships.
- operational AI and workflow automation;
- internal tools that connect teams, data, and existing software;
- agent-assisted delivery with human approval and visible failure paths;
- product discovery and rapid, testable prototypes;
- translating between business users and technical implementation.
This profile is a selected public surface, not a complete work timeline. Most production and client work remains private; I can share sanitized architecture and implementation walkthroughs where appropriate.