Skip to content

Work

Client data platforms since 2022, as a consultant with Modeo, Eulidia and Ippon Technologies.

Mar 2026 – presentvia Modeo

Accor

Procurement data platform

I work on every layer of the Snowflake platform of Accor’s procurement department: ingestion, Snowflake infrastructure in Terraform, dbt models on Airflow, exports and production support.

Tickets delivered since March, across ingestion, Terraform, dbt and exports.
40+
Migrations to dbt delivered, each cut over in production with no service interruption.
12
Daily production check, after I automated it.
45 to 5min
  1. Sources

    • About 15 systems
    • Files, SFTP, APIs
  2. Ingest

    • Python connectors
    • Schema checks
  3. Store

    • Snowflake
    • Terraform-managed
  4. Transform

    • dbt models, tests
    • Run by Airflow
  5. Serve

    • Dashboards, apps
    • Exports to CRM
    • and support tools

RunIncidents, fixes and follow-up with data providers. A daily automated check of ingestion, Snowflake and dbt results opens or closes Jira tickets.

What I built or changed

About fifteen source systems are ingested by Python connectors, validated, stored in Snowflake managed with Terraform, transformed by dbt on Airflow and served to dashboards, apps and operational tools. I work on every stage after the sources.
  • Snowflake
  • Terraform
  • dbt
  • Airflow
  • Python
  • SQL
  • n8n
  • Streamlit

Context

The data team of Accor’s procurement department runs a Snowflake platform fed by about fifteen source systems: suppliers, contracts, spend, invoicing and hotel data. It feeds the department’s dashboards and apps, and sends data back to operational tools such as the CRM and customer support.

What I do

  • Ingestion: build and maintain the Python connectors that pull data from SharePoint, Coupa, Zendesk and other sources, and the schema validation every file goes through before it reaches Snowflake. Built the hotel enrichment pipeline end to end, from an AI search API to Snowflake.
  • Snowflake and Terraform: manage the Snowflake infrastructure as code (schemas, tables, streams, tasks and roles) and the SQL transformations that still run as Snowflake tasks.
  • dbt migration: move those transformations to tested dbt models run by Airflow, one domain at a time (budget, invoicing for several regions, reference lists), each validated row for row against production before the cutover.
  • Data for the business: new tables, master data on hotels and contracts, and reporting exports for partners, plus the data sent back to the CRM and the support tool.
  • Production: incident investigation from the source file down to the dbt model, run fixes, follow-up with the providers who deliver the data, and code review for the team. I also automated the daily production check with n8n and store every dbt result in Snowflake.
  • Proofs of concept: Snowflake data served through an API with Snowpark Container Services, and work on the team’s Streamlit apps.

Jan – Mar 20253 monthsvia Eulidia

Effy

Semantic layer migration

Moved Effy’s dbt models to a MetricFlow semantic layer, so that every team queries the same metric definitions.

  1. Store

    • BigQuery
  2. Transform

    • dbt models
    • Run by Airflow
  3. Define

    • MetricFlow
    • Semantic models
    • Shared metrics
  4. Use

    • Self-service analytics
    • Reports

What I built or changed

Data in BigQuery is transformed by dbt models run by Airflow, exposed through a MetricFlow semantic layer and used for self-service analytics.
  • dbt
  • MetricFlow
  • Airflow
  • BigQuery

What I did

  • Introduced a single semantic layer with MetricFlow on top of the existing dbt project.
  • Migrated the dbt models, defined the semantic models and standardized the key business metrics.
  • Teams could then build their own analyses on consistent metrics.

2023 – 202414 monthsvia Ippon Technologies

EDF Renewables

AWS data lab

Set up a central data lab on AWS for EDF Renewables, then led its migration from data.all to Amorphic.

  1. Sources

    • Data from several departments
  2. Lake

    • S3
    • Glue, Athena
    • AWS CDK
  3. Govern

    • data.all
    • then Amorphic
  4. Use

    • SageMaker
    • Analysts, data scientists

What I built or changed

Data from several departments lands in S3, is cataloged and queried with Glue and Athena, governed first by data.all and then by Amorphic, and used by analysts and data scientists.
  • SageMaker
  • Athena
  • S3
  • Glue
  • AWS CDK
  • data.all
  • Amorphic

What I did

  • Set up a centralized AWS data lab under the AWS Migration Acceleration Program, giving departments one place to access data.
  • Configured data.all with AWS services and built the first workflows and pipelines, then ran the full migration to Amorphic: data models, pipelines and integrations.
  • Wrote the training material and ran onboarding sessions for technical and non-technical users.
  • Handled all English-language communication between EDF Renewables’ teams and the development team in London.

20233 monthsInternal project

Ippon Technologies

Salesforce and Snowflake integration

Led the technical build of a two-way integration between Salesforce and Snowflake.

  1. Source

    • Salesforce
  2. Ingest

    • Into Snowflake
  3. Transform

    • dbt models
  4. Deliver

    • Back into Salesforce
    • BI dashboards
Salesforce data is ingested into Snowflake, transformed with dbt, then sent back into Salesforce and used in BI dashboards.
  • Snowflake
  • dbt
  • Salesforce

What I did

  • Designed the data architecture, built the ingestion into Snowflake and the dbt transformations, and sent the processed data back into Salesforce.
  • Built BI dashboards on the unified data for business users.

2022 – 20233 monthsvia Ippon Technologies

Datanumia, EDF Group

Data platform rebuild

Replaced a chain of Python scripts on AWS Glue with a dbt project on Amazon Redshift.

  1. Sources

    • S3
    • Lambda
  2. Store

    • Amazon Redshift
  3. Transform

    • dbt models
    • dbt tests
  4. Document

    • Project documentation

ReplacedA chain of Python scripts running on AWS Glue.

What I built or changed

Files in S3 are loaded into Redshift, transformed and tested with dbt, then documented. The dbt project replaced Python scripts on AWS Glue.
  • Redshift
  • dbt
  • Glue
  • Lambda
  • S3
  • Python

What I did

  • Redesigned the data architecture and configured the Redshift cluster.
  • Rewrote the transformations as dbt models and owned their testing.
  • Wrote the full project documentation.

20226 monthsInternal project

Ippon Technologies

Cloud data platform

Built a cloud data platform from open-source tools during my six-month internship at Ippon.

  1. Source

    • PostgreSQL
  2. Ingest

    • Airbyte
  3. Store

    • Amazon Redshift
  4. Transform

    • dbt
    • Run by Airflow
  5. Serve

    • Superset dashboards

PlatformTerraform for the infrastructure, Keycloak and AWS Secrets Manager for access.

PostgreSQL data is ingested by Airbyte into Redshift, transformed with dbt and shown in Superset dashboards, on infrastructure provisioned with Terraform.
  • Redshift
  • dbt
  • Airbyte
  • Superset
  • Airflow
  • Terraform
  • Keycloak
  • Docker

What I did

  • Provisioned the infrastructure with Terraform, ingested PostgreSQL data into Redshift with Airbyte, transformed it with dbt and served it through Superset dashboards.
  • Secured the platform with AWS Secrets Manager and Keycloak.

Sketches