CUSTOMER STORIES
Coinbase Simplifies ML Workflows and Reduces Deployment Time from Months to Hours
With Snowflake ML capabilities,Coinbase is delivering machine learning at scale, eliminating complex pipelines and accelerating time to value.
Industry
Financial ServicesLocation
San Francisco, CAStory highlights
- Improving workflows for data scientists from months to hours: Snowflake ML transformed the machine learning process into an efficient workflow, eliminating the need to send data from Snowflake to other platforms like Databricks. This simplified the pipeline, allowing data scientists to perform feature engineering, model training and deployment in a matter of hours instead of months.
- Enhancing productivity and governance: By having everything in a single, seamless platform on Snowflake, Coinbase's team could be much more productive. This unified approach eliminates concerns about security and governance, providing data scientists with a smoother, well-governed and scalable workflow.
- Accelerating business outcomes: This better workflow leads to more timely and actionable insights for the business. Coinbase now can personalize products, improve customer retention and make customers happier with the platform — wins for data scientists and the business, too.
Video Transcript
Coinbase_TuhinGhosh_synced.mp4 [00:25:57 - 00:26:02]
Speaker 3: I'm Tuhin Ghosh. I lead up data science for the Platform Product Group at Coinbase.
Coinbase_TuhinGhosh_synced.mp4 [00:26:12 - 00:26:30]
Speaker 3: we are the team that's responsible for onboarding users, making sure payment systems making sure our blockchain ecosystem works. And we log down the platform against bad actors. So it's a very broad scope, and data plays a key role in everything that we do.
Coinbase_TuhinGhosh_synced.mp4 [00:07:38 - 00:07:54]
Speaker 3: So we have always used Snowflake for our analytical workloads. And we love the Snowflakes platform for what it gives us in terms of governance and scalability. And now we are able to leverage the platform to do sophisticated machine learning
Coinbase_TuhinGhosh_synced.mp4 [00:08:47 - 00:09:18]
Speaker 3: So our existing setup used to be that we would need to send the data, which oftentimes happens to be in Snowflake, on a little bit of a field trip. So we'd have to send it from Snowflakes to Databricks where the rest of the pipeline live. And that pipeline had a bunch of tools like Tecton and Fiddler, you name it. It was the alphabet soup. But what that meant was that for a data scientist to get a machine learning model up and running was a bit of a side quest.
Coinbase_TuhinGhosh_synced.mp4 [00:17:54 - 00:18:09]
Speaker 3: So with Snowflake ML, now they can do feature engineering, feature refreshes, model training, model deployment, refreshes of those models, and they can it in a matter of hours,
Coinbase_TuhinGhosh_synced.mp4 [00:27:27 - 00:27:40]
Speaker 3: Having everything in one place, and play seamlessly. Where you don't have to worry about security and governance, having everything in one platform just makes it such a productivity enabler,
Coinbase_TuhinGhosh_synced.mp4 [00:22:05 - 00:22:13]
Speaker 3: So it's a win on both sides. For data scientists, what this means is that they have a smoother, well governed and scalable workflow.
Coinbase_TuhinGhosh_synced.mp4 [00:10:32 - 00:10:46]
Speaker 3: as you can imagine, like having access to simple libraries and SQL specific ways of doing things like forecasting Coinbase_TuhinGhosh_synced.mp4 [00:10:50 - 00:10:55]
Speaker 3: This used to take months, we are turning that into days, if not hours.
Coinbase_TuhinGhosh_synced.mp4 [00:22:13 - 00:22:27]
Speaker 3: For the business, what it means is they have more timely, actionable insights which they can use to personalize the product, help retain our customers and make our customers overall happier with the Coinbase platform.
