Thought Leaders in Artificial Intelligence: Comet ML CEO Gideon Mendels (Part 2)

Sramana Mitra: Do you go to market as a platform company?

Gideon Mendels: I guess it depends on your definition. The Community Edition plays a big role as individual contributors get excited about it and recognize the value it can bring to them and their team.

Sramana Mitra: I noticed that you are calling it the Community Edition but not an open source product. It is a free product but not open source?

Gideon Mendels: Correct. About a week ago, we launched an open source product which is more focused on debugging model predictions and datasets. The core experimentation and monitoring platforms are not open source, but they’re completely free.

The main reason is there is a significant overhead running a product like Comet if you think about ingesting all the data. It’s challenging to run it on someone’s laptop. That said, the majority of our enterprise customers are running Comet in a VPC.

Sramana Mitra: But you do support a significant community edition that is free and hosted on your infrastructure, right?

Gideon Mendels: Correct.

Sramana Mitra: What percentage of that free user base is converting into paid customers?

Gideon Mendels: We deal with teams and not individuals. In the Community Edition, it’s individual contributors. In the enterprise edition, it’s teams. A significant amount of our enterprise customers had some exposure to the Community Edition at some point.

Sramana Mitra: Let’s look at some use cases. What is your product being used for? What problems is it solving?

Gideon Mendels: It’s a developer tool similar to how you may think of something like GitHub. We are completely agnostic to what type of model you are building and what use case you’re trying to solve. If you think about it from the software engineering lens, it’s similar. You can write code and use GitHub to version control it. It’s very similar.

If you look at our customers, we do have coverage across all these industries – media companies, tech companies, e-commerce, automotive. We support a very wide array of use cases. When you look at it from the lens of ML ops, it does get more specific. Generally speaking, we haven’t met a customer building a model where we say that there is not a lot of value we can provide.

Sramana Mitra: Double-click down and tell me what is that value.

Gideon Mendels: Before I double-click, you can think of it almost as a system of records software where you have GitHub for software engineers, Salesforce for sales people, Workday for HR. Like a lot of these systems, the value proposition spans multiple personas.

If you think about the individual contributors like data scientists and machine learning engineers, they’re using Comet for tracking all their experiments. When you run a bunch of models, you want to come back to work tomorrow and look at what you did yesterday. You want to be able to reproduce your work.

If you have a model that works well, you want to be able to retrain it with new data. You have to know all the different pieces that affect that result. They’re looking into comparing and explaining all these experiments and digging in the results.

This segment is part 2 in the series : Thought Leaders in Artificial Intelligence: Comet ML CEO Gideon Mendels
1 2 3 4 5