11/10/2026 updated
100 % available
Freelance Cloud & Data Engineer — Google Cloud Certified (Professional ML Engineer)
Modena, Italy
Worldwide
Bachelor's Degree in Computer EngineeringAbout me
Freelance Cloud & Data Engineer specialising in Google Cloud, with 3 years building and running production data systems. Based in Italy, working fully remote across Europe. Italian native, English C1. Available immediately.
Application Programming Interfaces (APIs)Apache AirflowGoogle BigQueryCloud ComputingCloud DatabaseCloud EngineeringData ArchitectureGoogle AnalyticsPython (Programming Language)PostgreSQLMachine LearningMongoDBPublic CloudCloud ServicesReverse Engineering
Every technology below is one I have run in production. Where and how, project by project, is in my project history.
GOOGLE CLOUD. BigQuery: the warehouse layer on both my main projects. Staging, curated and golden-record models on the MDM migration, and the target of the MongoDB to BigQuery migration that cut query times by around 90%. Partitioning, clustering, tuning, and the reconciliation queries that validated a cutover record by record. Dataflow: high-volume batch transformation over 15+ million customer records, and the pipelines turning raw marketing feeds into clean datasets across 50+ client accounts. Pub/Sub: the event backbone of the MDM platform, with source change events driving downstream processing in place of scheduled batch ETL, plus dead letter topics and retry policies. Cloud Run: containerised services running MDM business logic (record matching, survivorship rules, golden records) and REST APIs consumed by downstream enterprise applications. Cloud Functions: Pub/Sub and Cloud Storage file-arrival triggers, plus glue steps between stages. Cloud Workflows: orchestration across services, with sequencing, conditional branches, retries and error handling. Secret Manager: credentials and API keys with automatic rotation, read by service accounts, no secrets in code. Cloud SQL (PostgreSQL): operational persistence layer. Cloud Build: Git-based CI/CD for container images.
AIRFLOW AND DOCKER. Airflow: production DAGs running daily ingestion across 50+ client analytics accounts, with task dependencies, scheduling, retry policies and backfills for multi-year analyses. Docker: an image per Cloud Run service, and standardised execution environments across teams.
LANGUAGES AND DATA. Python daily. SQL at an advanced level: window functions, complex CTEs, tuning. Pandas, NumPy, FastAPI. PostgreSQL, BigQuery, MongoDB, MySQL, Firestore.
WHAT I AM USUALLY HIRED FOR. Migrating legacy data platforms to cloud-native architectures, including reverse engineering of undocumented Informatica PowerCenter ETL. Building and operating ingestion and transformation pipelines. Data quality and reconciliation against source systems. REST API design with OAuth2.
AWS. Lambda, S3, DynamoDB and CloudWatch, in production for two years. Certification: Google Cloud Professional Machine Learning Engineer, January 2026.
Experience
Professional experience0 - 5 years
Freelancer since2026
Languages
EnglishFluentFrenchBasic knowledgeItalianNative speakerSpanishBasic knowledge
Project history
MakerDojo brings coding, robotics and AI into Italian schools. For its teacher training project I design and deliver a hands-on generative AI programme for teachers of a public school network in Piacenza, from preschool to middle school. Teachers leave with materials they built themselves and know how to check.
What I do
• Designed the curriculum from scratch at MakerDojo's request, then reworked it around the school's constraints.
• Adapted it to four subject areas (language, maths, inclusion and neurodiversity, arts and music) and three school levels, each with its own examples, prompts and one extra tool vetted for cost, age limits and terms of use.
• Lead four-hour in-person workshops: how language models work and where they fail, a guided demo, then long labs on teachers' own materials.
• Produce facilitator guides, slide decks, ready-to-use prompts and a feedback survey for every session.
• Built rules on minors' data into every session: no student data in any tool.
Tools: Google Gemini, NotebookLM, Canva for Education, Teachable Machine, Goblin Tools, Quick, Draw!, Suno
Highlights
• Curriculum delivered within a 38-session programme for 19 teacher groups, shared with a second trainer.
• A five-part prompting framework (role, context, task, format, constraints) that non-technical teachers apply on their own.
• Second AI training engagement after a full-day programme for senior staff at Tecnomat.
Supporting and integrating international students as part of the Erasmus Student Network in Italy.
Sole engineer on the migration of a legacy Informatica PowerCenter MDM platform to an event-driven serverless architecture on Google Cloud, for a major Italian multi-brand retail group: 8 brands, 15+ million customer records.
How each technology was used:
Pub/Sub: the event backbone. Change events from the source systems drive downstream processing, which replaced scheduled batch ETL with near real-time reaction. Dead letter topics and retry policies.
Cloud Run: containerised services running the MDM logic (record matching, survivorship rules, golden records) and the interfaces consumed by downstream applications.
Cloud Functions: Pub/Sub and Cloud Storage file-arrival triggers, plus glue steps between stages.
Cloud Workflows: orchestration of the migration and the recurring pipeline: sequencing, conditional branches, retries, error handling.
BigQuery: staging, curated and golden-record models, plus the reconciliation queries comparing new output against the legacy Informatica output record by record before cutover. Partitioning, clustering, tuning.
Dataflow: high-volume batch transformation across the full customer base.
Secret Manager: Cloud SQL (PostgreSQL) credentials and API keys, automatic rotation, read by service accounts.
Docker and Cloud Build: an image per Cloud Run service, Git-based CI/CD.
Also reverse engineered undocumented PowerCenter mappings and defined the cutover strategy.