Pretzel AI

Pretzel AI

Open-source & AI-native replacement for Jupyter Notebooks

**Jupyter is used by millions** of data scientists and analysts but it has _many_ problems. **Pretzel fixes Jupyter's problems** with native AI-tooling, collaboration, modern code-editing features, native SQL support and a visual, AI-assisted data exploration and dashboarding suite. **We're two ex-Stripe founders**. Prasoon worked as a Data Scientist at Stripe, Goldman Sachs and Klarna for over 7 years. We recently launched on GitHub and got **1000 GitHub stars in two weeks**, 500+ DAU, usage at large companies like Stripe and Bayer and 7 new code contributors.

Pretzel AI
Team Size:2
Location:Berlin, Germany
Group Partner:Dalton Caldwell

Active Founders

Prasoon Shukla

Prasoon worked at Stripe, Goldman Sachs and Klarna for over 7 years as a Data Scientist, building ML models and analytics for Stripe's payment fraud, Klarna's US credit market and Goldman's market risk data. He graduated from IIT Roorkee in India with a degree in Applied Maths.

Prasoon Shukla
Prasoon Shukla
Pretzel AI

Ramon Garate

Ramon was an Engineer at Stripe and Klarna and was the first Lead at 2 early stage startups. First line of code in SaaS (startups), releasing new products to consumers (Klarna app) and big integrations with Enterprise (Stripe). He holds a MSc in Autonomous Vehicles Dynamics and Control from Cranfield University.

Ramon Garate
Ramon Garate
Pretzel AI

Company Launches


Hey everyone! We're Prasoon and Ramon from Pretzel AI, and we're building the modern Jupyter Notebook that fixes many longstanding problems with Jupyter. Pretzel improves on Jupyter with native support for AI tools like GitHub Copilot and LLMs, built-in collaboration (think Google Docs), an IDE-like code-writing experience (think VSCode), native support for SQL and Python, reactivity (so your notebooks are reproducible) and a visual data exploration and dashboarding suite. Best of all, we are fully open-source.

As our first product, we’ve built just one feature Pretzel: a visual data exploration tool that runs locally in the browser. See demo below or check us out on GitHub!

❌ The Problem with Jupyter Notebooks

Jupyter Notebook started in 2011 as a nicer, interactive way to work in Python. At the time, it was the best interactive tool to work with data really popular! It’s currently used by millions of data scientists and analysts everyday.

Despite it’s popularity, there are many longstanding problems with Jupyter:

  1. Notebooks do not reproduce reliably. Running cells out of order, running them multiple times, and hidden state can lead to inconsistent results.
  2. Collaboration is difficult. There's no easy way to work on a notebook with others in real time like you can with Google Docs.
  3. The code-writing experience is subpar compared to modern IDEs like VSCode. Features like intelligent autocomplete, linting, debugging, GitHub Copilot support and in-place editing with LLMs are all lacking.
  4. Integrating with other tools and languages can be clunky. While Jupyter supports kernels for other languages, you can't seamlessly switch between SQL and Python natively.
  5. Notebooks are still only usable by technical folks - i.e., data scientists and analysts. Most non-technical people still prefer using visual tools like Metabase or Tableau for their data needs.

🥨 The Solution: Introducing Pretzel

We're building Pretzel to address these pain points by building the notebook experience for the modern day. Here's what makes Pretzel special:

  1. AI-Powered: Pretzel integrates natively with AI tools like GitHub Copilot and large language models. Get intelligent code suggestions, understand, edit, and generate code right from your notebook.
  2. Reproducibility: Pretzel notebooks are reactive, meaning that cells update automatically when their dependencies change. This ensures your notebooks always stay in sync and produce consistent results.
  3. Collaboration: With Pretzel, you can collaborate on notebooks in real-time with others, just like in Google Docs. No more emailing notebooks back and forth or dealing with merge conflicts.
  4. IDE-like Experience: We're bringing the best parts of modern IDEs to Pretzel. Enjoy intelligent autocomplete, linting, debugging, testing, and more right in your notebook.
  5. Polyglot: Pretzel has first-class support for both SQL and Python. Switch seamlessly between languages in the same notebook.
  6. Open-Source: We believe in the power of open-source software. It eases adoption significantly (vs paid tools), has strong network effects and best of all, users can build their favorite features and contribute back to the community!

💡 What Are We Doing Differently?

Other tools like Hex have tried to fix problems with Jupyter. While Hex is a great product, it has been unable to displace Jupyter as the de facto data tool. Why? Because Open Source = Distribution and closed source tools have failed at distribution.

With Pretzel, we tap into the same distribution mechanics as Jupyter to become the new de-facto data tool. With our free and open-source approach:

  • Way more people are able to try Pretzel for their work because it's free and self-serve - it can be installed just like Jupyter to run on your local machine without needing to worry about procurement cycles and data-sharing concerns.
  • Network Effects come into play. When people contribute code into Pretzel to solve their problems, everyone benefits. The more people that use Pretzel, the more it’s value.
  • Enterprises are more willing to try Pretzel because they're not worried about us going out of business in two years!

We plan to monetize via cloud hosting and enterprise features (like SSO, compute, single-tenancy hosting, local LLM support etc).

📣 Ask

We'd love to get your feedback on Pretzel and the features you would like to see! Email us at founders [at] withpretzel [dot] com or talk to us on GitHub.

Here’s a silly photo of us to wrap up the post :)

Happy data wrangling!

Prasoon and Ramon, Co-Founders of Pretzel AI