Hamed ZITOUN

Freelance mission through NeoSmartDayta, on the Data Platform team. • Rebuilt a legacy data platform (Sqoop, Rancher) into a Data Mesh on Airflow 3 and AWS Kubernetes (EKS). • Migrated Airflow 2 to Airflow 3 one DAG at a time: two instances side by side, rollback in one file. • Built a self-service data platform: a business team adds a new source to Snowflake with one YAML file. The Airflow tasks, the load and the quality checks are created for them. • Moved data freshness from 2pm to 4am. Business teams get their data ten hours earlier every day. • Cut Snowflake storage cost by half: cost alerts, detection of unused tables and clones, query tags to see the cost of each pipeline. • AI Champion for the data team: Claude Code agents, skills and routines for incident triage, code review, job scaffolding and onboarding. • AI-assisted operations: agents read the logs, tell a real failure from a transient one, and give the on-call engineer a ready diagnosis in Slack. • Data quality and governance as code: validation rules, drift detection on Snowflake roles and grants, pseudonymisation of personal data, RBAC and Vault policies. • Real-time CDC pipeline: critical tables refresh every 30 minutes. • 13 sources ingested (Stripe, Jira, Anaplan, Pigment, web scraping...), dbt and Snowflake transformations, activation to Braze, Amplitude and Hightouch, BI in Power BI and Microsoft Fabric. • CI/CD with GitHub Actions, Docker, ECR, ArgoCD and Terraform. Monitoring with Prometheus and Grafana. Stack: Airflow 3, Snowflake, dbt, Python, AWS, Kubernetes, Terraform, GitHub Actions, Claude Code, MLflow, Arize.

My own consulting company in AI and Data Engineering. All my freelance missions are delivered through NeoSmartDayta (https://neosmartdayta.com). • I take clients from the first need to a system in production on AWS and GCP: data platforms, ETL/ELT and big data pipelines, RAG and multi-agent apps, and ML pipelines. • Hands-on on every mission: MLOps and LLMOps, LLM evals, guardrails and token cost tracking, monitoring with Arize and W&B, and a clear focus on business results. Contact: [email protected]

Freelance mission through NeoSmartDayta. • Built Telemetry, the central real-time data platform of Talend, so the Data Science and Analytics teams get clean and trusted product usage data. • Collected events from 5 regions (EU, US, APAC, AU, Azure): 5 Kafka clusters, 25 Debezium CDC instances, 25 MongoDB replicas and 25 PostgreSQL clusters, all streamed into one central AWS MSK cluster. • Delta lakehouse (Databricks) with Bronze, Silver and Gold zones. 5 Spark Streaming jobs clean and normalise the data. The Gold zone in Snowflake feeds the Tableau dashboards. • Built churn prediction models on product usage data (SparkML, XGBoost, scikit-learn). • Led the team for 7 months as Principal Data Engineer. Worked with the SRE team on FinOps and cost cuts. Stack: Kafka, Debezium, Spark, Databricks, Kubernetes, AWS (EKS, MSK, EMR, S3, Athena, Glue), Snowflake, Python, Java, Helm, Flux, Grafana, Prometheus.

• Built a real-time ETL from MongoDB to BigQuery with Airflow on Cloud Composer. It reads MongoDB oplogs, maps events to BigQuery, handles failures and restarts from a checkpoint in Cloud SQL, with alerts in Cloud Logging. • Automated BigQuery access control with Cloud Functions and Pub/Sub: new tables are visible only to the right users and roles. • Synced business metadata from CSV files into BigQuery. • CI/CD with GitLab on a custom Kubernetes runner, Terraform. Stack: GCP (BigQuery, Composer, Cloud Functions, Pub/Sub, Cloud SQL), Airflow, Python, MongoDB, Docker, Kubernetes, Terraform.

• Won an innovation challenge run by Banque des Territoires for local authorities with two partners, then co-founded the startup autogeo.ai (https://autogeo.ai). • Built a deep learning model that georeferences water network plans automatically: it places an AutoCAD plan (PNG or PDF) on the cadastral map. Stack: CNN, autoencoders, FAISS semantic search, OpenCV, GIS and GeoJSON, Flask, Docker, GCP.

• Built a Python API that recognises veterinary medicines from a photo, so Ceva can track how its products are used. • Pipeline: object detection and OCR with Google Vision API, image preprocessing, then fuzzy text matching and a combined score to pick the right product. Stack: Python, Google Vision API, Elasticsearch, Cloud Storage, Cloud Run, Flask, Docker.

• Built an NLP matching engine that suggests the right fix procedure for each support ticket, so Carrefour support agents solve tickets faster. • Owned it end to end: pre-sales, PoC and client demo, then the production module (spaCy, RoBERTa embeddings, FAISS, MARISA trie, ranking, precision and recall benchmarks). • Shipped as a REST API with Flask, Docker and Kubernetes on GCP.

• Built Vidal Sentinel, a real-time web app that scores the drug risk of each hospital patient, so pharmacists know which prescriptions to check first. The app won the France HIT Innovation Award 2019. • Designed the scoring model (scikit-learn, served with Flask) and a reactive REST API (Vert.x, RxJava, MongoDB, Hazelcast, RabbitMQ, FHIR). • Built the Angular front end. TDD, BDD and pair programming.

Client: Direction Générale des Entreprises (DGE). IoT R&D for Intelligent Transport Systems and Smart Cities. • Joined as an intern, then grew to Tech Lead and Scrum Master in 4 years. • Led the front-end and back-end teams of a real-time web portal (Angular, Spring, WebSocket, ActiveMQ, MQTT, Docker, Jenkins). • Built the native Android app for the R&D use cases, a oneM2M node talking MQTT to the IoT platform (OAuth 2.0, NFC, maps).

✔️Course 1 summary: - Container, Docker, and Red Hat OpenShift architecture - Create containerized services - Manage containers and container Images - Create custom container images - Deploy containerized applications on Red Hat OpenShift - Deploy multi-container applications ✔️Course 2 summary: - Install an OpenShift Platform Container cluster - Configure and manage masters and nodes - Secure OpenShift Platform Container - Control access to resources on OpenShift Platform Container - Monitor and collect metrics on OpenShift Platform Container - Deploy applications on OpenShift Platform Container using Source-to-Image (S2I) - Manage storage on OpenShift Platform Container