How Activeloop uses Bugout to ethically build Data 2.0, future of data for machine learning purposes
Activeloop is a YC company whose mission is to help data scientists seamlessly connect their unstructured data to machine learning models. Activeloop’s open source library, Hub, has over 3200 stars and was recently featured as the top Python project on GitHub.
The Activeloop user community is growing rapidly, doubling month over month over the past two quarters.
GitHub featuring us on their Trending page brought a lot of users into our community and made us realize that we didn’t know which parts of Hub people were using or what issues they faced.
We suspected that, for every issue that our users raised on GitHub or in our Slack, there were something like 15 other issues going unreported.
We care deeply about the Hub developer experience and wanted to know about all the issues that our users were facing.
What we needed was an analytics solution that could tell us how our users were using Hub. Web applications use Mixpanel and Segment for this, but Hub is a Python library and those tools aren’t well suited for our purpose.
In our search for a solution, Bugout initially stood out because of the respect with which it treats the end user and privacy measures it brings to the table.
Bugout’s reporting library, humbug, collects the information we need to effectively solve problems for our users - stack traces, system information, and so on. humbug also makes it easy to define a consent flow in which we present our users with the option to opt in or opt out of reporting in a principled manner.
Integrating Bugout into our code base was very simple. It introduces very little overhead in terms of code and performance.
The speed with which we were able to get up and running with Bugout made it our clear choice for analytics.
We use Bugout in two ways:
To collect anonymous statistics on how often users are using Hub.
To collect crash reports when a user runs into an exception while using Hub.
We’re confident that by clearly understanding our users' behavior on their local machines, we can design a solution that works for everyone, on any machine, at scale.
Bugout provides us with critical insights. The data from Bugout informs our product roadmap, helping us prioritize the improvements to the Hub code base. Importantly, we can evaluate the exact problems our customers are facing. We also get a full view of how our Open Source package is being used, while being respectful of each individual user's privacy.
Davit Buniatyan
CEO at Activeloop