How MoneyLion saves hours weekly on analytics using Deepnote
MoneyLion’s data team runs weekly analytics and exploratory analysis for multiple stakeholders. Previously, the team was operating on local machines and JupyterHub, with unstable sessions and no native AI assistance. With Deepnote, they work in a collaborative notebook workspace connected to Snowflake, share results as dashboard-style apps instead of slides, and move from questions to insights in minutes with built-in AI.
Switching from JupyterHub to Deepnote took almost no effort. We plugged directly into Snowflake and immediately spent less time troubleshooting kernels and deployments, and more time on delivering actual insights.
Melvin Low
MLOps platform engineer
SQL to chart in seconds
with built-in visualization and text to SQL
Saved ~2 hours per analyst per week
by avoiding chart setup and reruns
Apps instead of slides
to share results with stakeholders
Saved ~8 hours per month per headcount
by retiring JupyterHub Helm chart maintenance
Challenge
Local and JupyterHub limits slowed the work
Local laptops lacked memory and CPU for large datasets, while JupyterHub sessions were unstable and could disconnect, forcing reruns and lowering efficiency. As analyses grew, JupyterHub became laggy, and some cells would not render. Additionally, there was no native built-in AI coding assistance, which contributed to visualization work taking a long time.
Sharing and reproducibility were clunky
Stakeholders did not like the previous sharing flows. Links often required a manual run to see charts and tables, and it was hard for non-authors to inspect logic or queries. The process of taking results from code output and moving them into presentation slides for weekly meetings was manual and prone to error. Every EDA started from scratch, requiring repetitive copy and paste for similar analyses, which was their biggest blocker to faster answers.
Platform maintenance added overhead
Maintaining public JupyterHub helm charts pushed a search for a better notebook environment. Databricks was evaluated, but it is a full platform rather than a focused notebook solution, so the team chose a best-in-class notebook, Deepnote.
Deepnote’s AI handles the repetitive tasks, like rerunning SQL queries and setting up exploratory analyses. That means we get more time to rapidly scale and collaborate on experiments.
MoneyLion
data science team manager
Solution
Plug and play across the stack
Deepnote connects directly to Snowflake, and [please add other integrations you are using] for analytics and to other stores through preferred clients and SDKs. Access follows Snowflake roles, similar to their previous approach. The team can start a fresh notebook, attach the needed integrations, and get moving.
AI data visualization
Analysts move from query to chart in seconds. They can use built-in AI for text to SQL and visualization to get the first view, then refine with drag and drop fields, chart type, axes, and legends, all in the same notebook.
Apps replace slide copy
Custom apps provide dashboard-like behavior for weekly reviews, so stakeholders view live results without manually copying outputs into slides.
Results
Faster analytics with fewer steps
The MoneyLion team now creates charts quickly, which lowers the time from question to first insight. Combined with stable, long-running cells that no longer need supervision, weekly reviews move faster, and analysts keep focus on the analysis rather than restarts and re execution.
More reliable execution and fewer do-overs
Large notebooks now run without the lag and rendering issues the team saw in JupyterHub, which keeps reviews and investigations moving.
Less friction for stakeholders
Stakeholders can inspect SQL logic and see charts without hitting Run, which reduces back and forth and rebuild requests for the weekly huddle.
Lower platform overhead
The platform team moved away from maintaining public JupyterHub helm charts for stability. Deepnote’s managed environment removes that recurring work.
Deepnote gets us from raw SQL to polished visuals instantly, so we can focus on delivering real business value, rather than tweaking charts.
MoneyLion
data scientist
What’s next
MoneyLion plans to expand from EDA and weekly analytics to custom model training modules in Deepnote. They can package training code as Modules, pull features from Snowflake, pass run parameters through an app, and schedule jobs on a cadence. Results, metrics, and plots would be written back to Snowflake or object storage and surfaced in lightweight apps for review. Templates and parameters make runs reproducible, secrets and Snowflake roles keep access aligned with policy, and Git keeps code changes traceable.