Marimo: Reactive Python Notebooks for Reproducible & Fast Data Workflows

Marimo positions itself as next-generation Python notebooks — reactive, reproducible, and engineered for modern data workflows. In this session, Bartosz Mikulski, our Senior ML Engineer, demonstrates how Marimo addresses long-standing issues in Jupyter-based environments: merge conflicts, hidden global state, non-deterministic execution, fragile cell order, and the constant need for throwaway helper code. What you’ll learn: ⚙️ how reactive execution and DAG-based state remove hidden state issues 📄 why Python-backed notebooks make version control and merging painless 📊 how to build dynamic UI with sliders, dropdowns, data previews, and interactive charts 🚀 how Marimo integrates with Polars, DuckDB, and SQL for fast analytics 🌐 how to turn your notebook into a clean, stakeholder-ready web app 🧠 caching, async, AI-assisted code generation, embeddings exploration (Fashion-MNIST, LEGO dataset) If you work with Python notebooks daily, this walkthrough offers a grounded look at tools designed for reproducible, maintainable, engineer-friendly workflows. 00:00 Intro & Agenda 03:19 Marimo vs. Jupyter 06:35 Demo of Marimo 13:01 Data Exploration in Marimo 28:20 Final Thoughts Check our website: https://deepsense.ai/ Linkedin: https://www.linkedin.com/showcase/applied-ai-insider #Marimo #Python #ReactiveNotebooks #ReproducibleDataScience #NotebookReproducibility #Polars #DuckDB #AIEngineering #ModernDataTools #JupyterAlternatives