Why Sumble replaced Jupyter with marimo, from notebooks to apps

Anthony Goldbloom (Founder and CEO Sumble, formerly Founder and CEO Kaggle) says, “There have been a few technologies that have been truly transformational for me…marimo is one of them. Our company is dramatically better instrumented because of marimo, and it’s our gateway into many AI-enabled applications. Today, just about everything you need to know about Sumble lives in marimo.”

marimo is world’s best development environment for AI and data, and it’s all free and open source. Just pip install marimo or uv add marimo to get started.

Sumble is building a knowledge graph for the world’s external data. Today, they provide AI-enabled account intelligence data, enabling sales teams to do deep research. Companies large and small use Sumble to better inform targeting, outreach, and lead enrichment. With a 15-person team spanning engineering, go-to-market, and data science, Sumble’s mission is to make external data as searchable as the internal data companies rely on.

Challenge

As a fast-growing, product-led business, Sumble needed a rapid, flexible way to build interactive data tools and dashboards for both internal use-cases and customers. Before marimo, Founder/CEO Anthony Goldbloom and his team faced numerous problems using Jupyter and a long tail of other tools to create data applications at his former company Kaggle (acq. by Google) and in the early days of Sumble:

1. Painful, clunky app development: Anthony recounts “A simple React-based data app took me two weeks to build and share internally. Even with other, ‘modern’ tools, it was a painful, 8 hour+ process.”
2. Fragmented workflow: As a heavy notebook user, Anthony was most comfortable in an environment that allowed for interactive computing, but would need to constantly shuttle between Jupyter and other platforms when creating data apps, manually translating and updating whenever changes were needed
3. Maintenance overhead: Customer-facing dashboards required constant updates, but the back-and-forth between Jupyter notebooks and data app development frameworks slowed the team down and led to tons of throwaway analyses and brittle data apps that were ultimately too hard to maintain. Jupyter notebooks are intended to be scratchpads, not reusable software artifacts, and this led to constant challenges translating from notebook to application
4. Limited usability: Traditional data apps created by engineering and data science rarely reached GTM, meaning that sales and marketing team members were often left without real-time insights and stale, static dashboards instead. An organization that prided itself on being data-driven was left scrambling to keep operational teams up to date

Sumble needed a way to quickly turn notebook explorations into production-ready, low-maintenance applications that could be deployed across their entire team. Enter marimo.

Using marimo

Sumble was ecstatic to adopt marimo, an open-source development environment that feels like a next-gen, AI-native notebook, but is stored as Git-friendly Python that can be pushed to production. marimo was the obvious platform to build the data tools, AI applications, and dashboards Sumble desperately needed.

marimo transformed Sumble’s development process almost overnight. Today, marimo allows the company to go from prototype/notebook -> dynamic data app instantly. “What took days in React takes basically no time in marimo.” Anthony says. Whenever Anthony has a customer question, he can instantly spin up a reusable application on the spot with marimo.

As the number of data applications grew, so did the cross-functional stakeholders who needed access to marimo apps. As sales reps came onboard, the Sumble team containerized their data apps with Docker and is now investing in migrating to cloud hosting for seamless, cross-team access. marimo now powers 26 internal applications for Sumble. The use-cases range from account-scoring dashboards, real-time usage metrics, to company-wide reporting. Sumble is now creating more data apps, faster than ever and is being accelerated by the unique combination of Claude Code and marimo. Anthony believes “the addition of marimo plus Claude Code will mean that we write apps that we wouldn’t otherwise.”

Sumble is also using marimo to build AI-driven internal applications including automating outbound email campaigns and an AI-native resume screener specific to Sumble’s candidate pipeline and funnel. According to Anthony, using marimo, a code-first platform, is critical in the age of AI because it lets him use APIs of the latest models like Gemini to create powerful bespoke tools like their resume screener. marimo powers custom software creation, at scale.

Anthony says, “I used to write dozens of throwaway notebooks or hard to maintain data apps. Today, all those notebooks are reusable marimo applications… Kaggle at its best never had analytics as good as Sumble, and that’s because of marimo.”

Results

marimo has evolved from a simple prototyping tool into Sumble’s system of record for analytics and internal applications. By embracing marimo, the team has realized four key benefits:

1. Faster and more flexible data app creation: From two weeks in React -> hours in other environments -> instantly in marimo. marimo allows Sumble to answer customer questions instantly and spin up complex, interactive data tools/dashboards on the fly.
2. Seamless cross-functional reporting: marimo is now a source of truth for Sumble’s go-to-market team efforts. The entire sales team uses marimo apps daily, and engineers are using the tool multiple times a week for prototyping and deeper analytics.
3. Significantly reduced technical debt and no ongoing maintenance of critical data apps: Dashboards and internal data tools require almost no upkeep, as all marimo notebooks are intended to be reusable software artifacts out of the box.
4. AI-native innovation: marimo’s seamless integration with agents like Claude Code has empowered Sumble to build sophisticated, AI-driven internal tools that were previously out of reach for them.
5. Dramatically better data work. In addition to using marimo to make apps, Sumble’s engineers and data scientists use marimo notebooks to interactively explore and transform data on a weekly basis; marimo’s reactive execution, built-in dataframe explorer, and SQL integration make it far more useful than legacy notebooks like Jupyter.

Founder and CEO Anthony Goldbloom says, “There have been a few technologies that have been truly transformational for me … marimo is one of them. Our company is dramatically better instrumented because of marimo, and it’s our gateway into many AI-enabled applications. Today, just about everything you need to know about Sumble lives in marimo.”