Software Engineer · Rome

Spina means plug in Italian. Hence MichaelPlug.

Go and Python backends, data lakehouses and tokenization platforms, built for systems that can't go down.

Portrait of Michele Spina
shipping software since 2021
5+ years
from aerospace to film, detailed below
9 projects
peer-reviewed, CAiSE 2024 and 2025
2 papers

About

Since 2024 I have been a Software Engineer at Moveax in Rome, designing and building Go and Python services for high-availability environments, following Hexagonal Architecture and Clean Code principles: data lakehouses on Iceberg and Trino, financial tokenization infrastructure on Polygon, and mission-critical APIs.

From 2021 to 2024 I worked as a freelance engineer, delivering native macOS applications and data automation tools for international film productions.

Alongside my master's degree I collaborated with research groups at Sapienza, Utrecht University and TU Munich on confidential data sharing over public blockchains, co-authoring two papers published at CAiSE 2024 and 2025.

education
MSc, Computer ScienceSapienza, Rome · 2024 · 110/110 cum laude
BSc, Computer ScienceSapienza, Rome · 2021
languages
Italian, English

Experience

The parts of a system people only notice when they stop working. That's where I've spent most of my time.

Software Engineer @ Moveax

Mar 2024 – Present · Rome

Engineering backend systems, data platforms and Web3 architectures using Hexagonal Architecture principles.

Aerospace data lake

Data Governance Platform

A company-wide data catalog built on OpenMetadata: what data exists, who owns it and who may see it.

  • Setup as code: a Python tool that builds domains, teams and access rules from YAML.
  • Visibility: people see only their own branch of the domain tree.
  • Interface: a rebranded UI and new pages to follow data ingestion.
Illustration of a domain tree: a business unit with two divisions and three departments, each level with its own team: Data Managers, Business Owners, Data Owners. A user in department A1 sees their department, its division and the business unit. The other branches stay hidden.
  • OpenMetadata
  • Python
  • Java
  • React

Aerospace data lake

Data Access Platform

One controlled way into the data lake for every application. Admins choose which tables and columns each app can read, and every query is checked and logged.

  • Policies: turned into OPA rules that Trino checks on every query.
  • Credentials: app keys from creation to reset, with role checks.
  • Audit: every query logged with the reason behind it.
Illustration: an application asks for a flights table. The policy check lets the query through, and the result comes back with three allowed columns while the pilot column stays locked. The query is written to the audit log with the app's name and its business reason.
  • Python
  • FastAPI
  • Trino
  • OPA
  • React

Aerospace data lake

Flight Data Dashboard

Dashboards where a helicopter maker's engineers and customers explore recorded flights, one at a time or across a fleet. I built much of the FastAPI backend behind them.

  • Charts: altitude, maneuvers and fleet stats, queried with Trino from the data lake.
  • Access: customers only see their own helicopters.
  • Exports: large downloads run by a background worker.
Illustration of a dashboard card: the altitude of one flight over time, split into climb, cruise, hover and descent. Labels above the chart show that the view is limited to the customer's own fleet and can be exported.
  • Python
  • pandas
  • FastAPI
  • Trino
  • Iceberg
  • Redis

Ticket Sales Network

The platform behind a regional bus operator's ticket sales network: shops, partners and on-board staff selling tickets, each within a budget, with revenue split on every sale. I worked across the full stack, mostly on the Go services, building audit logging, account deletion with personal-data obfuscation, background workers and tracing from gateway to Sentry.

Diagram in four layers. Mobile apps, web back-offices and partner integrations call NestJS gateways over GraphQL or REST. The gateways call Go microservices over gRPC, which run the business logic, background workers and audit trail. They store data in Oracle, feed dbt analytics, and call an external ticketing provider.
  • Go
  • gRPC
  • Oracle
  • NestJS
  • GraphQL
  • React Native
  • dbt

Bond Tokenization

A platform for issuing regulated bonds as tokens on Polygon. Each euro payment and its token transfer settle together or not at all. I ran the research phase: twelve design records, from proxy patterns and network choice to custody and finality, most backed by a working proof of concept. Then I built Go services for minting, investor settlement, document notarization and refunds, plus a Hardhat plugin that signs deployments with AWS KMS.

Diagram of a bond's life in four steps: a factory issues the bond and its coupon tokens, investors subscribe, coupons are paid each period, and at maturity the principal is refunded and the tokens burned. A close-up of the subscribe step shows the investor's euro payment and the issuer's bond tokens entering a hash-locked escrow from opposite sides. The escrow releases both to the other party together, or returns both.
  • Solidity
  • ERC-3643
  • Polygon
  • Go
  • Kafka
  • AWS KMS
  • Hardhat

Synthetic Data Exchange

Built the consent layer of a health-data sharing platform. Providers' sharing policies are recorded on Polygon by a Solidity contract, then decide which consumers, public or private, can download the synthetic datasets generated from their data.

Diagram in two steps. First, a provider's consent goes through the TypeScript backend and is stored as a policy on Polygon. Then that policy decides which synthetic datasets public or private consumers can download from S3.
  • Solidity
  • Polygon
  • Hardhat
  • ethers.js
  • Fastify
  • S3

Energy Forecasting

Built the Dagster pipelines behind weekly consumption forecasts for buildings. Every floor and resource (electricity, hot and cold water) runs in parallel: it fetches its history, forecasts the next four weeks with a SARIMA model and stores the results in TimescaleDB.

Diagram: a weekly Dagster schedule fans out into one parallel run per floor and resource. Each run fetches history, forecasts with SARIMA and stores a 28-day forecast in TimescaleDB.
  • Python
  • Dagster
  • TimescaleDB
  • Alembic
  • Kubernetes

Freelance Software Engineer @ CineDev

Jul 2021 – Feb 2024 · Remote

Engineered native macOS applications and data automation tools for major international film productions.

Camera Metadata Manager

Main developer of Altera, a native macOS app that crews on set and in post use to manage camera-clip metadata, on productions for Marvel and Sony. I wrote about 70% of its commits as it went from first beta to the 1.0 release, on a 50k-line Swift codebase, and owned the release lifecycle.

Diagram: camera and editorial files are ingested, organized into bins and layouts, then delivered as editorial exports and reports.

I built the interchange with Avid and other editing tools, the export engines, and a template language crews use to design their own reports. I also brought the app to Apple Silicon and kept it shipping through macOS updates.

  • Swift
  • AppKit
  • Xcode
  • Python

Film Research Automation

Built a Python tool that turns any IMDb search into a ready-to-use spreadsheet for CineDev's market research. A simple CSV decides what to collect for each film, such as camera crew, writers, countries or box office, so non-developers can change the report without touching code.

TitleCameraWritersBox office
titlecamera dept: key grip, gripwriters (written by)budget, worldwide gross
The whole config is one CSV: name a column, then list the IMDb fields that fill it. Each film in the search becomes a row.
  • Python
  • pandas

Publications

Public blockchains copy every record to every node. That's great for trust and bad for secrets. These two papers, written with researchers at Sapienza, Utrecht and TU Munich and published at CAiSE, encrypt documents so the ledger proves they exist while only the right people can read them.

MARTSIA

A Tool for Confidential Data Exchange via Public Blockchain

CAiSE 2025 · arXiv:2407.10684

Spreads trust across several independent authorities. Each one issues part of a decryption key and the reader puts the parts together, so no single party can unlock the data. One export document can be split so customs, the carrier and the manufacturer each read only their own part: one message instead of ten copies.

Runs on Ethereum and Algorand, and was tested with three existing apps for NFT trading, supply chain and retail.

  • Python
  • Solidity
  • PyTeal
  • IPFS
  • Ethereum
  • Algorand
CAKE

Sharing Slices of Confidential Data on Blockchain

CAiSE 2024 · arXiv:2405.04152

Encrypts a document, or separate parts of it, with a rule about who may read it, such as "customs on shipment 29837". The encrypted file goes to IPFS and its fingerprint goes on-chain. Only readers who match the rule get a working key.

Used in BRIE, a national cybersecurity project for import-export logistics.

  • Python
  • Solidity
  • PyTeal
  • IPFS
  • Algorand
  • Docker

Skills

The tools I reach for, and where I've actually used them.

Backend

Data Platforms

Web3

Frontend & Apps

Storage

Infra & Ops

Shameless plug: tell me about it, and I'll make it real. Email is the quickest way to reach me.