I study and build across the entire stack — from Rust backends, AI agent workflows, and distributed systems to web interfaces — figuring out what is worth building with the people who use it.
I'm a software engineer in Bologna driven by curiosity: whatever a problem requires — AI agents, distributed Rust backends, web and mobile interfaces, or custom engines — I study it deeply and build it end to end. At Luna S.r.l. I design and ship production software with clients, sitting with them to decide what is actually worth building before writing code. On my own time I build the developer tools and simulations I was missing, led by Trama and Mnemosyne. I graduated from the University of Bologna in 2026 with a thesis on real-time telemetry (PETRA).
I build MCP servers, deterministic guardrails for LLMs, and multi-actor coordination tools where humans and coding agents ship together.
Built Trama's Rust MCP server, Mnemosyne's semantic code memory for AI, and TrainGame's validated LLM narrator.
Event-driven architectures, async stream consumers, type-safe ORMs, and end-to-end observability in production environments.
Designed a client's production event-driven Rust platform at Luna S.r.l. and published ducklake-orm on crates.io.
Fast, accessible interfaces and operator consoles designed around real workflows, live WebSocket state, and clean visual systems.
Shipping operator consoles and a ~200-vehicle fleet platform in production, plus Trama's full web workspace.
Server-authoritative multiplayer, content-addressed storage daemons, and deterministic ECS worlds compiled to native or WebAssembly.
Authoring Eivar-Online's predicted multiplayer simulation and Mnemosyne's always-on local daemon.
At Luna S.r.l. I work across client software: a fleet platform with Java/Spring Boot, Angular and React Native; an event-driven Rust platform I designed for a client; and a new operator console with Spring Boot APIs and a React interface. Before I implement, I talk the options through with the client: what to build now, what can wait, and what each choice costs the people running the platform. The code is reviewed by a senior.
Tecnologie dei Sistemi Informatici
Design and Development of PETRA, an Event-Driven Platform for Real-Time Asynchronous Telemetry and Analysis
Real systems built from scratch: human + AI agent coordination, AST semantic history, a published Rust ORM on crates.io, full-stack platforms, and multiplayer Rust simulations.
The missing layer between a reported problem and a shipped fix. Issues stay. A workstream carries the outcome, with the decisions and the artifacts attached.
Issue trackers assume one ticket is one person and one pull request. That falls apart when several bugs share a cause, the work crosses repositories, and a coding agent does part of the job. The reason for a choice ends up in a chat nobody can find a month later.
1 / 9 Multiplayer prototype: server-authoritative Rust simulation, replicated state, client-side prediction.
Local-first semantic history for source code: symbols and structural changes via Tree-sitter, not line diffs — queryable by AI agents over MCP.
Rust ORM for DuckDB and DuckLake, published on crates.io — derive macros, query builders, pooling, migrations, time travel.
Self-hosted Albion guild platform — bank, loot splits, battle analytics, Discord login.
A life sim on an endless train: hundreds of NPCs with jobs, a closed economy and gangs, and an LLM that proposes one novelty a day, which the engine checks before it becomes real.
First I understand who uses the software and what each option costs. Then the problem becomes a verifiable workstream in Trama, I build it in testable increments, review every diff line by line, and leave clear context for the team.
Before writing code, I understand who needs it, how they work and what a useful result looks like. The stack comes after the context.
The problem becomes an issue in Trama before it becomes code. Then I decide what belongs in the change, what can wait, and how to know it works.
Connect the pieces, not just the layers: data, logic and interface. Build in small increments that can be run, tested and understood.
Question the implementation, test the edge cases and make each diff easy to review. Writing code is only part of the work; checking it matters just as much.
Explain what changed, why and how to verify it. A decision that should last goes on the workstream in Trama, not only in a chat.
The fastest way to reach me is email — I read everything and reply.
