Thousands of tools exist. The right reflex is to organise them by use: analyse, store, transform, visualise, deploy, govern.
3.1Getting started: the student kit
- Spreadsheets: Excel or Google Sheets remain the world's number one data tool.
- SQL: a universal skill, required in almost every data job offer.
- Python and pandas: the standard for analysis and data science, often in Jupyter or Google Colab.
- Open data: data.gouv.fr, data.europa.eu, Kaggle to practise on real data.
3.2Databases, warehouses and lakehouses
- Databases: PostgreSQL, MySQL, SQLite, MongoDB, Redis.
- Vector databases: pgvector, Pinecone, used for semantic search and RAG.
- Warehouses / lakehouses: Snowflake, BigQuery, Databricks, Redshift.
- Open formats: Parquet, Iceberg, to limit vendor lock-in.
3.3Transformation, orchestration, real time
dbt transforms data with versioned SQL; Airflow orchestrates pipelines; Spark processes distributed volumes; Kafka handles real-time streams.
3.4Visualisation and BI
Power BI, Tableau, Looker, Metabase turn tables into dashboards. Golden rule: a good chart answers a question; a bad one raises ten.
3.5The cloud in practice
- Hyperscalers: AWS, Microsoft Azure, Google Cloud.
- French players: OVHcloud, Scaleway, Outscale, Clever Cloud.
- Modern tooling: Docker, Kubernetes, Terraform, GitHub Actions.
3.6AI in the data stack
Data analysis assistants, NotebookLM, text-to-SQL, BI copilots: AI quickly executes what it is asked to do, mistakes included. It therefore makes a real command of data even more valuable.
How to choose?
Three questions: what volume and what speed? What sensitivity? What reversibility? Favouring open standards (SQL, Parquet, containers) keeps an exit door open.
Sources
Documentation of the tools cited; CNIL, reference frameworks and the PIA software (cnil.fr); data.gouv.fr, DINUM/Etalab.
Finished this module?
Your progress is saved on this device.