Thought Leaders in Artificial Intelligence: Comet ML CEO Gideon Mendels (Part 1)

We’ve covered ML Ops before in interviews such as Arize.

Gideon provides a comprehensive overview of how the space is evolving and the opportunities on the horizon.

Sramana Mitra: Let’s start by introducing our audience to you as well as to Comet.

Gideon Mendels: I’m the CEO and Co-Founder of Comet. I actually started my career as a software engineer about 16 to 17 years ago. I shifted to working on machine learning about nine years ago as a grad student where I mostly worked on language modeling. At that time, it was mostly focused on speech recognition systems. Now it’s all over the place.

After that, I joined Google and worked on hate speech detection on YouTube comments. My shifting from software engineering to machine learning is the biggest reason for starting Comet. When we think about the software world, we have this amazing stack of tools like databases, testing, orchestration, and versioning. Each one of those has multiple-billion dollar companies. Then you join a machine learning team and someone emails you a Jupyter notebook. This was back in 2016.

I called my co-founder, Nimrod, who I’ve worked with in the past. After a brief conversation, it was obvious to us that there’s an opportunity to do so much better here. We started Comet. Essentially we power these data science and machine learning teams. These are platforms to manage the entire development workflow of machine learning models all the way from tracking the different datasets used to train these models to tracking the experiments themselves.

Machine learning, unlike software engineering, is a research-driven paradigm. You have to try many things before you find something that works. You end up running all these experiments with different datasets, algorithms, and hyperparameters until you find a model that satisfies the metric that you’re optimizing for. We provide a lot of tooling around that to analyze these different experiments, comparing and understanding why one is doing better than the other.

Eventually, once these teams are successful in bringing a model to production, we also help monitor these models’ performance and production. Whether it’s some downstream issue with feature corruption or more traditional form of concept drift where the universe around the model changes, you need to be able to detect that.

Quite naturally, if the universe around the model changes, the performance would degrade. We provide a platform that detects these changes, alerting around them, and helping those teams find the root cause analysis.

To give you a bit of background about the business, we have two sides. We have a community edition that is 100% free. Everyone can go to comet.com and get a free account. That’s used by tens of thousands of data scientists around the world. Then we have our commercial offering. We support the best machine learning teams in the world. These are companies like Etsy, Zappos, Uber, and Shopify.

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