
Romain Courivaud engineered and maintained the MTES-MCT/zero-logement-vacant analytics platform over 15 months, delivering robust data pipelines and scalable analytics features. He designed and refactored ETL workflows using Python, SQL, and dbt to support evolving housing vacancy analytics, integrating external data sources and automating ingestion into DuckDB and MotherDuck. His work included deploying Dagster for orchestration, enhancing data validation, and modernizing Docker-based environments for reliable CI/CD. Romain improved data quality through rigorous testing, schema migrations, and documentation, enabling accurate business intelligence and machine learning workflows. His contributions demonstrated depth in data engineering, cloud deployment, and analytics infrastructure.

February 2026 Monthly Summary – MTES-MCT/zero-logement-vacant
February 2026 Monthly Summary – MTES-MCT/zero-logement-vacant
January 2026 monthly summary for MTES-MCT/zero-logement-vacant: Focused on delivering two key features and validating deployment workflows. Major bugs fixed: none reported this month. Overall impact: produced production-ready analytics tooling to explore community indicators and exit rates, improved housing vacancy exploration and ML training capabilities, and established cloud deployment workflow for faster iterations. Technologies demonstrated: Streamlit, data exploration, data visualization, ML training, Python, code refactoring, deployment to Clever Cloud, Git-based version control.
January 2026 monthly summary for MTES-MCT/zero-logement-vacant: Focused on delivering two key features and validating deployment workflows. Major bugs fixed: none reported this month. Overall impact: produced production-ready analytics tooling to explore community indicators and exit rates, improved housing vacancy exploration and ML training capabilities, and established cloud deployment workflow for faster iterations. Technologies demonstrated: Streamlit, data exploration, data visualization, ML training, Python, code refactoring, deployment to Clever Cloud, Git-based version control.
December 2025 monthly summary for MTES-MCT/zero-logement-vacant focused on delivering robust data ingestion, analytics improvements, and more accurate entity lookups. The work hardened the data pipeline, improved data quality, and expanded geographic/demographic context for housing vacancy analyses.
December 2025 monthly summary for MTES-MCT/zero-logement-vacant focused on delivering robust data ingestion, analytics improvements, and more accurate entity lookups. The work hardened the data pipeline, improved data quality, and expanded geographic/demographic context for housing vacancy analyses.
November 2025 performance summary for MTES-MCT/zero-logement-vacant: Delivered a scalable external data ingestion framework and owner events analytics, enabling direct loading of external sources into DuckDB/MotherDuck, centralized configuration for public data providers, and standardized asset naming for analytics (dbt owner_events). Implemented a robust external pipeline to build external sources and refreshed dbt schemas. Stabilized pipelines with infrastructure and environment improvements across Dagster and Metabase Docker images, including Python 3.12 upgrades and necessary build dependencies. These efforts improved data reliability, scalability, and timeliness of owner-related analytics, reducing operational risk and enabling faster, data-driven decisions. Technologies demonstrated: DuckDB/MotherDuck, dbt, Dagster, Docker, Python 3.12, ETL architecture and data modeling.
November 2025 performance summary for MTES-MCT/zero-logement-vacant: Delivered a scalable external data ingestion framework and owner events analytics, enabling direct loading of external sources into DuckDB/MotherDuck, centralized configuration for public data providers, and standardized asset naming for analytics (dbt owner_events). Implemented a robust external pipeline to build external sources and refreshed dbt schemas. Stabilized pipelines with infrastructure and environment improvements across Dagster and Metabase Docker images, including Python 3.12 upgrades and necessary build dependencies. These efforts improved data reliability, scalability, and timeliness of owner-related analytics, reducing operational risk and enabling faster, data-driven decisions. Technologies demonstrated: DuckDB/MotherDuck, dbt, Dagster, Docker, Python 3.12, ETL architecture and data modeling.
Monthly summary for 2025-10: Delivered foundational data platform improvements for MTES-MCT/zero-logement-vacant, focusing on production reliability, developer ergonomics, and expanded analytics capabilities. Key features delivered include: - Dagster development environment and dependency updates to streamline local setup and avoid installation issues; - Added marts_productions_events table to enhance event tracking and data warehousing; - Simplified marts_production_events data model by removing joins to marts_production_join_owner_housing and owner_id casting; relying on a direct join with marts_production_users to reduce complexity; - Consolidated ff and lovac schemas into a single cerema schema to simplify ingestion and maintenance; - Production data tests validation improvements to ensure reliability in production. Major bug fixes include: - Production housing analytics tests corrected by fixing dbt schema test configuration and renaming tests to data_tests, with refined accepted values and relationships to prevent production failures. This reduced false positives and improved data quality in production.
Monthly summary for 2025-10: Delivered foundational data platform improvements for MTES-MCT/zero-logement-vacant, focusing on production reliability, developer ergonomics, and expanded analytics capabilities. Key features delivered include: - Dagster development environment and dependency updates to streamline local setup and avoid installation issues; - Added marts_productions_events table to enhance event tracking and data warehousing; - Simplified marts_production_events data model by removing joins to marts_production_join_owner_housing and owner_id casting; relying on a direct join with marts_production_users to reduce complexity; - Consolidated ff and lovac schemas into a single cerema schema to simplify ingestion and maintenance; - Production data tests validation improvements to ensure reliability in production. Major bug fixes include: - Production housing analytics tests corrected by fixing dbt schema test configuration and renaming tests to data_tests, with refined accepted values and relationships to prevent production failures. This reduced false positives and improved data quality in production.
September 2025 summary for MTES-MCT/zero-logement-vacant: Delivered three core items that improve data analytics, data fidelity, and deployment reliability. Implemented Housing Vacancy Analytics with NLP-enabled data processing and anonymization tooling, including a fix to campaign return-rate reporting. Added Production Events Mart with owner/IDs to improve event data provenance and consistency. Modernized Dagster environment and Docker tooling with pyproject.toml and uv.lock, and hardened Dockerfile practices to ensure reliable dependency installation and builds.
September 2025 summary for MTES-MCT/zero-logement-vacant: Delivered three core items that improve data analytics, data fidelity, and deployment reliability. Implemented Housing Vacancy Analytics with NLP-enabled data processing and anonymization tooling, including a fix to campaign return-rate reporting. Added Production Events Mart with owner/IDs to improve event data provenance and consistency. Modernized Dagster environment and Docker tooling with pyproject.toml and uv.lock, and hardened Dockerfile practices to ensure reliable dependency installation and builds.
Monthly summary for 2025-08: Focused on expanding analytics capabilities and data model clarity for MTES-MCT/zero-logement-vacant. Implemented notes data ingestion, launched a comprehensive vacancy exit analytics suite, and improved data models documentation. The changes strengthen production analytics pipelines, governance, and usability of data products. No major bugs fixed this month; improvements were delivered with updated documentation and configuration to support scalable analytics pipelines.
Monthly summary for 2025-08: Focused on expanding analytics capabilities and data model clarity for MTES-MCT/zero-logement-vacant. Implemented notes data ingestion, launched a comprehensive vacancy exit analytics suite, and improved data models documentation. The changes strengthen production analytics pipelines, governance, and usability of data products. No major bugs fixed this month; improvements were delivered with updated documentation and configuration to support scalable analytics pipelines.
In July 2025, MTES-MCT/zero-logement-vacant delivered four core outcomes that enhance data quality, analytics consistency, and the robustness of the analytics stack. Key features and fixes improved data accuracy and insight depth, while a modernization of the analytics environment streamlines maintenance and delivery of business insights.
In July 2025, MTES-MCT/zero-logement-vacant delivered four core outcomes that enhance data quality, analytics consistency, and the robustness of the analytics stack. Key features and fixes improved data accuracy and insight depth, while a modernization of the analytics environment streamlines maintenance and delivery of business insights.
June 2025: Delivered batch BAN addresses population and analytics integration for MTES-MCT/zero-logement-vacant. Implemented batch processing to update BAN addresses and auto-populate missing owner addresses; added a Dagster asset and job for missing BAN addresses; integrated analytics imports, and improved logging. This work enhances data completeness, analytics readiness, and observability, enabling faster business decisions around vacant housing data. Minor workflow refinements included removing ban score validation and updating the job name to reflect updated scope. Commit surface includes: 4e5664a90f8b57a8f16956caff4d20e5df53ed60; 93d16aee9b253eccfd10e9604b44de223231c1c2; 231c41145a7824da97124f5dbd522c88c91d3305; 85c0800082d686e4e32f3b0003e485af885286ac; 6cb165ead47cc9dda04c5ec0ae24783a29966f11; 0c59b6187ee881d49eae794b47065130518b2d39; 4ed0218627ecf148cd757be673c07b2c7da2bc13.
June 2025: Delivered batch BAN addresses population and analytics integration for MTES-MCT/zero-logement-vacant. Implemented batch processing to update BAN addresses and auto-populate missing owner addresses; added a Dagster asset and job for missing BAN addresses; integrated analytics imports, and improved logging. This work enhances data completeness, analytics readiness, and observability, enabling faster business decisions around vacant housing data. Minor workflow refinements included removing ban score validation and updating the job name to reflect updated scope. Commit surface includes: 4e5664a90f8b57a8f16956caff4d20e5df53ed60; 93d16aee9b253eccfd10e9604b44de223231c1c2; 231c41145a7824da97124f5dbd522c88c91d3305; 85c0800082d686e4e32f3b0003e485af885286ac; 6cb165ead47cc9dda04c5ec0ae24783a29966f11; 0c59b6187ee881d49eae794b47065130518b2d39; 4ed0218627ecf148cd757be673c07b2c7da2bc13.
May 2025 performance summary for MTES-MCT/zero-logement-vacant. Delivered major data ingestion enhancements and naming consistency that tightened the analytics pipeline, improved data quality, and accelerated time-to-insight. Key features were designed with 2025 readiness in mind and focused on simplifying the data processing stack for maintainability and scalability.
May 2025 performance summary for MTES-MCT/zero-logement-vacant. Delivered major data ingestion enhancements and naming consistency that tightened the analytics pipeline, improved data quality, and accelerated time-to-insight. Key features were designed with 2025 readiness in mind and focused on simplifying the data processing stack for maintainability and scalability.
March 2025 monthly summary for MTES-MCT/zero-logement-vacant focusing on delivering analytics features, robust ingestion, and deployment improvements. Key outcomes include expanded establishment data classifications, 2024 FF ingestion pipelines, Metabase/DuckDB reliability enhancements, data quality analytics capabilities, and cleanup of legacy notebooks. These efforts have driven deeper insights, more reliable dashboards, and streamlined data ops for the business.
March 2025 monthly summary for MTES-MCT/zero-logement-vacant focusing on delivering analytics features, robust ingestion, and deployment improvements. Key outcomes include expanded establishment data classifications, 2024 FF ingestion pipelines, Metabase/DuckDB reliability enhancements, data quality analytics capabilities, and cleanup of legacy notebooks. These efforts have driven deeper insights, more reliable dashboards, and streamlined data ops for the business.
February 2025 monthly summary for MTES-MCT/zero-logement-vacant: Delivered foundational data-geography enhancements and hardened deployment to support more reliable analytics and faster iteration cycles. The month focused on delivering key features, fixing critical infrastructure issues, and elevating data quality for public mart analytics and housing metrics.
February 2025 monthly summary for MTES-MCT/zero-logement-vacant: Delivered foundational data-geography enhancements and hardened deployment to support more reliable analytics and faster iteration cycles. The month focused on delivering key features, fixing critical infrastructure issues, and elevating data quality for public mart analytics and housing metrics.
January 2025 focused on stabilizing the data platform and delivering production-ready analytics capabilities for MTES-MCT/zero-logement-vacant. Key outcomes include establishing DWH synchronization, deploying Dagster and Metabase to the main environment, and expanding DuckDB/MotherDuck data ingestion. Substantial bug fixes improved container orchestration and reliability (Dagster port and Docker fixes), while security hardening and governance updates reduced risk. The month also delivered enhanced export and DBT workflows, setting the stage for scalable, observable analytics and faster business insights.
January 2025 focused on stabilizing the data platform and delivering production-ready analytics capabilities for MTES-MCT/zero-logement-vacant. Key outcomes include establishing DWH synchronization, deploying Dagster and Metabase to the main environment, and expanding DuckDB/MotherDuck data ingestion. Substantial bug fixes improved container orchestration and reliability (Dagster port and Docker fixes), while security hardening and governance updates reduced risk. The month also delivered enhanced export and DBT workflows, setting the stage for scalable, observable analytics and faster business insights.
December 2024 monthly summary for MTES-MCT/zero-logement-vacant. Focused on delivering data infrastructure enhancements to support housing vacancy analytics. Key work centered on establishing event status tracking and consolidating housing data, with refactoring of analytics queries to improve data processing and reporting latency.
December 2024 monthly summary for MTES-MCT/zero-logement-vacant. Focused on delivering data infrastructure enhancements to support housing vacancy analytics. Key work centered on establishing event status tracking and consolidating housing data, with refactoring of analytics queries to improve data processing and reporting latency.
November 2024 performance summary for MTES-MCT/zero-logement-vacant focused on delivering a major enhancement to housing occupancy and vacancy analytics and strengthening the data pipeline for cross-year analysis.
November 2024 performance summary for MTES-MCT/zero-logement-vacant focused on delivering a major enhancement to housing occupancy and vacancy analytics and strengthening the data pipeline for cross-year analysis.
Overview of all repositories you've contributed to across your timeline