↓ Skip to main content
Terraforming the Cloud Logo Logo Animated
Terraforming the Cloud

CV

1188 words
Table of Contents

Projects
#

Ambrogio — Personal Agent Orchestrator – Personal project
#

The problem: working with many agents — some piloted, others the terminals you normally use (Codex, Claude Code) — and keeping track of what they are doing: who works on what, how far they got, project by project, with one single place to look at the results and the repositories in order, without confusion.

How it solves it: Ambrogio is an AI-agent orchestrator and an audit trail in one binary. External agents are used natively and talk to it by working the tasks and producing the final documents — Ambrogio keeps the trail. For simpler jobs, or for any reason, Ambrogio is the one that creates a piloted agent to do the work. Site: ambrogio.tazlab.net.

How it is structured: projects, and inside each project the tasks become queues of tasks with a declared ceiling; the engine holds the chain and the trace — who, what, when, and why the queue was closed.

One central agent, one binary: nothing to install; internal cron jobs launch internal or external agents — internal agents read and search inside the engine, the external ones stay the terminals they always were; some cron jobs serve the system to keep everything in order, others are configured for routine jobs on the projects.

  • GitHub: the worker commits in its own worktree and delivers; the engine pushes the branch, opens the pull request, merges it after the check and brings the local trunk up — an agent never holds GitHub’s credential.
  • Sandbox and secrets: every agent runs in a Landlock sandbox (declared read and write paths, network and IPC restricted), clean and with a deadline: it receives only the secrets of its project (global + project), never the engine’s.
  • Projects, documents and the two doors: a project is a place with a root, documents are the texts an agent writes (plan, roadmap, status, revisions), deliveries are reports with verifiable artifacts. Static Go with SQLite inside, page, API and events served by the binary; the operator’s door behind a token, the agents’ door.

TazLab — Cloud-Native & AI Engineering platform – Personal project
#

Aug 2025 – present

A personal project that grew from a single Docker-container server into an enterprise-grade platform (~80 pods): the jump came from professional use of AI agents, employed to design, implement and maintain the whole infrastructure.

  • Kubernetes GitOps: multi-node Talos Linux cluster (Proxmox VMs) with Flux CD and Terraform/Terragrunt; one-shot bootstrap in ~12 minutes, with no manual work.
  • Fully IaC infrastructure: Proxmox, Terraform/Terragrunt (cluster, VMs, LXCs), Ansible (idempotent playbooks); every component can be destroyed and recreated in minutes, with data persistent on S3.
  • Secrets, PKI and zero-trust: HashiCorp Vault as the single backend (dynamic secrets, Vault Secrets Operator), 3-tier PKI with an offline Root CA, mTLS on PostgreSQL with automatic rotation — zero application passwords; Tailscale network as code with no public IPs.
  • Disaster recovery: offsite S3 backups, autonomous recovery after a real power outage (~10 min), full recovery from zero in <30 min.
  • AI Engineering: Mnemosyne (Go MCP server, pgvector semantic memory), Hindsight (multi-agent memory, 745+ memories), Ambrogio (agent orchestrator, Go); TazPod (Go CLI) as a workspace with AI agents on Proxmox VMs — one dedicated to the Hermes agent — reachable over SSH/Tailscale.
  • AI agents inside the process: Claude Code, Codex, OpenCode, Pi, Gemini and OMP within GitOps, CI/CD and multi-agent memory; Multica and Buzz as self-hosted control planes on Kubernetes. Documentation on wiki.tazlab.net and blog.tazlab.net.

Work Experience
#

Mathematics and Physics Teacher, Secondary School — Mantova – Mantova, Italy
#

Sept 2013 – present

  • Teaching & Education: theoretical and practical teaching of Mathematics and Physics, including curriculum design, assessment, and student preparation for state examinations.
  • IT Infrastructure Management: administration and maintenance of the school’s IT infrastructure, including the school website, electronic register, and Google Workspace platform.
  • AI Innovation & Education: design and delivery of training courses on the conscious, ethical, and effective use of Generative Artificial Intelligence for students and staff.

Programmer, Freelance – Bologna, Italy
#

Nov 2007 – Apr 2011

Web portal design and development.

  • Designed and developed complete web portals with dynamic front-ends and server-side interfaces for efficient content management.
  • Developed high-usability administrative interfaces to optimize portal management operations.
  • Applied technical skills: JavaScript, HTML, XML, Python, JSP, Servlet; SQL databases, RDF, OWL, Semantic Web.

Back-End Developer Analyst, Objectway – Milan, Italy
#

Nov 2003 – Nov 2007

Team back-end development (clients ING Bank and H3G).

  • ING Bank — Conto Arancio: contributed to the development and integration of back-end modules for the launch of the first branchless home banking service in Italy, helping with testing and debugging for the stability of the platform.
  • H3G: contributed to the development of a web configuration interface for pre-smartphone mobile phones and of one of the first push notification systems to customers; helped build back-end components able to handle traffic peaks of thousands of requests per second.
  • Technologies used: Java, JavaBeans, SQL databases.

Research Fellow, University of Bologna – Bologna, Italy
#

Sept 2001 – Nov 2002

Design and development of Artificial Intelligence systems for medical diagnostics.

  • Designed and implemented an expert system based on Neural Networks and Genetic Algorithms for automatic anomaly identification in mammographic images (early breast cancer diagnosis).
  • Developed predictive models (SVM, Neural Networks) for DNA Microarray analysis, aimed at identifying genetic markers and personalizing therapies.
  • Applied technical skills: C, C++, Java; Artificial Neural Networks, Genetic Algorithms, Support Vector Machine (SVM).

Education
#

SSIS Bologna, Teaching qualification (SSIS) in Mathematics and Physics – Bologna, Italy
#

2007 – 2009

CEFRIEL — Politecnico di Milano, Second-level Master’s degree in Information Technology – Milan, Italy
#

2003 – 2004

University of Bologna, Degree in Physics (110/110) in Physics — Electronics, Cybernetics & Artificial Intelligence track – Bologna, Italy
#

1995 – 2001

Languages
#

Mother tongue: Italian

English: B2

Skills
#

Platform Engineering & Cloud-Native: Proxmox VE as the base + fully IaC infrastructure: Terraform/Terragrunt (cluster, VMs, LXCs), Ansible (configuration and tooling installation, idempotent playbooks) and GitOps with Flux CD for the cluster; Kubernetes (Talos Linux), Docker/Podman, Helm, Kustomize, CI/CD (GitHub Actions); the whole environment can be destroyed and recreated in minutes, with persistent data on S3; distributed storage (Longhorn), observability (Prometheus, Grafana, alerting).

MLOps & AI Infrastructure: Containerized deployment of inference and semantic-memory services on Kubernetes; PostgreSQL + pgvector (embeddings, vector search); MCP protocol (server and client); end-to-end LLM operations (Hindsight, Mnemosyne): knowledge-base migration and audit, quota/rate-limit management, monitoring and recovery; orchestration of AI agents embedded in GitOps/CI-CD processes.

Security & Zero-Trust: HashiCorp Vault (PKI, dynamic secrets, Vault Secrets Operator), mTLS certificates with automatic rotation, gopass/GPG, Tailscale private network as code, cert-manager, Let’s Encrypt, Dex/OIDC.

Software Development: Go, Python, Bash (active); Java, C/C++ (previous experience); Git.

Cloud Providers: AWS (S3, IAM, SSO), Hetzner (in production), Google Cloud (Gemini API), Cloudflare, Oracle Cloud (OCI).

AI Agents — active professional use: Claude Code, Codex, OpenCode, Pi, Gemini, OMP and the self-hosted agent orchestrators Paperclip, Buzz and Multica: used daily to design, implement and maintain real infrastructure and processes, not just as coding assistants.

Comments