Turin, Italy.

I build things to find out what happens at the edges: when a dependency is slow, a peer disappears, a source file moves, a model changes, or a result needs to be explained later.

The repositories are the point. The short version is below; the details live in tests, fixtures, benchmarks, and docs.

Portfolio · Email · LinkedIn

package main

type Workbench struct {
    Current   string
    Questions []string
}

func main() {
    _ = Workbench{
        Current: "Go services, Rust experiments, and tools for reading systems",
        Questions: []string{
            "where did the latency go?",
            "what happens when this dependency is down?",
            "can the next person explain the result?",
        },
    }
}

4,655 contributions in the last year · 33 original public repositories · 31 stars across them · 53 followers

_{Updated weekly from GitHub's API. A profile snapshot, not a performance claim.}

• StreamHive: Go content-addressed storage over TCP/TLS, with framing, peer identity, replication, repair, durable stores, and metrics.
• Chatster: a real-time Go WebSocket system with SQLite history, bounded client queues, rate limits, metrics, and a React client.
• news-api: a TypeScript/Express service built around caching, upstream failures, request coalescing, OpenAPI, telemetry, and deployment.
• Quorabust: a Python duplicate-detection service that carries a model from training through serving, lineage, drift checks, and load testing.
• Forge: an installable Claude Code and Codex toolkit made of agents, skills, commands, policy hooks, and evals.

• STRATA · alpha: compiler-backed TypeScript architecture change control with evidence graphs, semantic diffs, and offline reports.
• LOOM · pre-alpha: local-first Rust/Tauri retrieval over text and Markdown, with SQLite FTS5, BLAKE3 hashes, and exact source anchors.
• CARTOGRAPH · pre-alpha: a TypeScript architecture graph scanner and Git revision diff tool that keeps evidence attached to relationships.
• GHOSTRACE · fixture-only: a local macOS causal event journal with explicit gaps, bounded provenance, and evidence-backed explanations.

The useful numbers are in behavior, outcomes, and shipped scope:

• Work: a MariaDB to PostgreSQL migration improved critical query performance by 3x to 10x and reduced infrastructure cost by about 60%.
• Chatster: the local fan-out harness delivered every message at 25 and 50 concurrent clients; p99 delivery latency was 5.7 ms and 57 ms respectively.
• LOOM: the synthetic retrieval fixture recovered all 3 expected sources at rank one with anchor precision 1.0.
• StreamHive: a 3-node Compose acceptance demo converges after durable restart and repairs a deleted blob through anti-entropy.
• Forge: 20 agents, 25 skills, and 22 commands packaged for Claude Code, Codex, and OpenCode, with policy hooks around the dangerous edges.

These are tied to a documented fixture, demo, or migration context; they are not promises about every environment.

I also work on private research around agent security, detection replay, evidence handling, post-quantum transport, network measurement, federated learning, privacy accounting, MPC, and traffic analysis. The names stay private until there is a result worth publishing.

I contribute to open source when the problem is concrete, including work around Go, databases, migration tooling, and developer tools.

Go · Rust · Python · TypeScript · PHP · SQL

PostgreSQL · SQLite · Redis/Valkey · AWS · Terraform · Docker · Kubernetes · Prometheus · OpenTelemetry

The list is intentionally incomplete. The repositories are a better description of how I work.