Arnaud Francois

WanData Project

- Context: WanData is an innovative SaaS platform, specifically designed to simplify the process of developing a data / AI strategy for local authorities and transport operators.

- Summary:

- KPI development: Implementation using GCS, Pandas, dbt, and BigQuery.

- Endpoint Exposure : Creation and serving of services via FastAPI.

- CI/CD Pipeline Setup : CI / CD using GitHub Actions, Artifact Registry, Podman.

- GCP Infrastructure Setup : Infrastructure design and deployment (Cloud Run, GCS, IAM permissions, Secret Manager, VPC, Terraform, Datadog, etc)

- Tool modernization : Migration of the dependency management tool (from Poetry to uv, from pylint to ruff).

- Participation in the modern data stack migration : Transition to a modern architecture (Airbyte, dbt, shifting indicators (KPIs) to OBTs (One Big Table)).

- PoC Gemini LLM agent using function calling. : Gemini / BigQuery / Jupyter / Streamlit

DataViz Project

- Summary: Analyzed trip launch rate and wild stops using data visualization.

- Technologies Used: GCS (Datalake), Airflow (Pipelines), BigQuery (DW), Tableau online

Modernization of Exports

- Context: Modernization of an export system aimed at reducing resource costs, preventing queue blockages, and minimizing support tickets. The exports consist of Excel or CSV files containing data on subjects such as trip histories, sales histories and so on. The system operates in a Serverless environment, with resources provisioned at the time of export.

- Technologies Used: Cloud Functions / Google Cloud Storage (GCS) / Python / Terraform / Cloud Build

API calls migration (modernization)

- Context: Transfer API calls data from the relational database to Pub/Sub and then an analytical database (BigQuery).

- Technologies Used: Pub/Sub, Dataflow (Apache beam), BigQuery, Java

Anonymization of usages

- Context: Anonymization of passenger usages to be GDPR Compliant.

- Technologies Used: Pandas, Airflow (Composer), Postgres

Next Trip Prediction Project

- Context: Designed and implemented predictive models for trip planning.

- Technologies Used: AI Platform, Google Cloud Storage (GCS), CSV, SKlearn, Weka, Jupyter Notebook. Usage Projection Project

- Context: Created usage projection models for optimizing transportation operations.

- Technologies Used: AI Platform, TensorFlow, LSTM, AutoKeras, Facebook Prophet (ARIMA/SARIMA). School Transportation Data Analysis

- Context: Collaborated on a study (ANATEEP) involving the consolidation of data from approximately twenty different transport networks. The objective was to analyze and provide valuable insights into school transport usage, focusing on indicators such as punctuality, service duration, and user-centric school runs.

- Technologies Used: Metabase, Snowflake (Data Warehouse), Talend (ETL).