I design, build, and deploy AI agents and product systems across the complete delivery path: model/runtime integration, tool execution, backend boundaries, web interfaces, infrastructure, CI/CD, observability, and release evidence.
Engineering portfolio · LinkedIn · Email
| Role | What I engineer | Core stack |
|---|---|---|
| AI Agent Engineer | Local and provider-backed agent runtimes, typed tool contracts, bounded execution, fallback, evaluation, memory/context flows, and traceable outcomes | Python, TypeScript, llama.cpp/GGUF, WebSocket/HTTP, Pydantic, provider adapters, deterministic test doubles, evaluation corpora |
| Platform Engineer | Reproducible environments, container delivery, multi-OS CI/CD, infrastructure automation, security gates, artifact provenance, health checks, and rollback paths | Docker/OCI, GitHub Actions, Linux, Windows, Terraform, Ansible, Helm, SBOM, dependency and secret scanning |
| Full-stack Engineer | Accessible web products with durable workflows, authenticated APIs, authorization, persistence, concurrency control, background jobs, and production-quality testing | TypeScript, React, Next.js, Python APIs, PostgreSQL/libSQL, MongoDB, RBAC, Playwright, accessibility and performance testing |
| Project | Engineering evidence |
|---|---|
| Cowork Prompt Enhancer | Released AI intent compiler with provider-independent execution, authenticated protocol boundaries, deterministic fallback, versioned evaluation, SBOM, provenance, and v1.0.0. |
| SysOps Agent | Public-alpha local operations agent with fail-closed classification, bounded commands, verification predicates, recovery loops, redacted audit trails, and green Windows/Linux gates. Host execution is explicitly non-sandboxed. |
| AutoBlog CMS | Full-stack editorial release candidate with database sessions, workspace RBAC, immutable revisions, conflict-safe editing, durable jobs, E2E, accessibility, security, and build evidence. |
| Authenticated Privacy Proxy | Containerized Mitmproxy–Privoxy–Tor chain with authenticated ingress, policy enforcement, DNS controls, host automation, and green CI/security scans; deployment security remains operator-dependent. |
- Claims include their environment, method, limitation, and reproduction path.
- Model output and external providers terminate at typed, tested, observable boundaries.
- Releases promote immutable artifacts with security, provenance, recovery, and rollback evidence.
Based in Rome, Italy; available for AI agent, platform, and full-stack product engineering work. This profile uses no tracking widgets, pixels, or generated activity claims.