Iterative has announced the launch of Iterative Studio Model Registry, the first machine learning model registry based on GitOps principles.
In this episode of TFiR Newsroom, Swapnil Bhartiya sits down with Dmitry Petrov, Co-Founder and CEO of Iterative, to discuss their latest offering, Iterative Studio Model Registry, and how its UI enables teams to see all the models and manage them across the ML lifecycle. He goes into detail about why open source plays such an important role for them and what the benefits are for going the open source route.
Key highlights from this video interview are:
- Iterative builds tools for machine learning based on open source. They recently released MLEM, a tool for model deployment and model registry. Petrov talks about the company’s offerings and their latest announcement of the Iterative Studio Model Registry.
- Petrov goes into detail about model registry, explaining that ML teams have been deploying their models to production in order to monitor the model and understand the styles. He explains what makes Iterative’s model registry different and how they are trying to apply the same practices from software development to the AI team.
- Having a source of truth, the best practice from the DevOps world is what sets Iterative’s model registry apart, according to Petrov. He tells us that a model registry is like a table in the cloud which says which model you have and the status of them. He discusses how their open source tools provide a UI to help visualize this and goes into detail about the other benefits of their model registry.
- Iterative is currently focusing on companies and teams inside the companies but does have the ability to share models publicly. However, Petrov explains that if your repositories are public then it is possible to share through the studio the status of the models but that this is not their current focus.
- Iterative is deeply committed to open source for multiple reasons. Petrov believes that OpenStack is the right way to build infrastructure and that this helps get fast feedback from customers and the community. He explains how this helps them improve their products and gain the trust of their customers.
- The goal of a model registry is to make the data science team and AI team responsible for the development models but also for the lifecycle of the models. Petrov feels there needs to be a breakdown of the silos between the UI team and the DevOps team and that this can be achieved by using the same set of tools and best practices. He explains who can benefit from Iterative Studio Model Registry and why.
Connect with Dmitry Petrov (LinkedIn, Twitter)
Learn more about Iterative (Twitter)
The summary of the show is written by Emily Nicholls.