Guest: Dmitry Petrov (LinkedIn, Twitter)
Company: Iterative.ai (Twitter)
Show: 2022 Prediction Series

Dmitry Petrov, Co-Founder & CEO of Iterative.ai, predicts that there will be emergence of new AI/ML tools as more and more AI/ML teams will be working with unstructured data including audio, video, images and unstructured text data. He also sees more specialization in the AI/ML field where people will focus on building expertise in certain areas vs unicorns who can do everything. What else did he predict? Check out in the video above.

[expander_maker]

Swapnil Bhartiya: Hi, this is your host Swapnil Bhartiya and welcome to our special series on predictions for 2022. Today, we have with us, once again, Dmitry Petrov, co-founder and CEO of Iterative. Dmitry, it’s great to have you back on the show.

Dmitry Petrov: Thank you for inviting me.

Swapnil Bhartiya: Before I ask to share your predictions, quickly tell us what the company is all about?

Dmitry Petrov: Company Iterative.ai, we build a machine-learning platform, but this is platform on top of software development stack. This platform utilizes all usual tools like GitHub, CICDs, and all the tools software engineers use in a database.

Swapnil Bhartiya: Excellent. Now it’s time for you to pick your crystal ball and share with us, what predictions do you have for 2022?

Dmitry Petrov: For 2022, we see a few trends. The first one is more and more teams. Machine-learning teams are working with unstructured data, which means images, audio, unstructured text data, and so forth and so on. People need to have some tool set for working with unstructured data, because with regular structured approach, there are tons of databases, database houses, but we need a new tool set for working with unstructured data, with video images, audio, et cetera. That will be a big trend this year. Another driver of this trend is data stand AI approach for machine-learning, which basically means instead of tuning models based on architectural change, some teams are tuning models on changing data, change labels. This is becoming a bigger and bigger trend. We expect more tools in foreign structured data management for label management, and so forth and so on.

The second trend is deeper specialization in the field because before, and many companies are still working on the kind of mode, when an engineer take care of the entire life cycle of the project from getting data, processing data, modeling, and then deploying this. That creates a lot of requirements for individual people, right? You need to have a unicorn who can do everything. Now, we are seeing that specialization is happening. There are some folks like ML engineers who are responsible for automation and resources, are focusing only mostly on the research part. That specialization is a clear trend right now.

Another driver for the trend is a lot of ML engineers and even resources like to contribute to infrastructure more than they get used to because complexity is changing, is shifting from the modern part to the calibration part to infrastructure part. Why it’s happening? Some deep learning improvements is in machine learning. It’s not as active as it was in 2017, 2018. The complexity move from the modern part to, basically, automation. Of course, the result of this is deeper specialization.

The third trend, which is related to the second one is collaboration because you have more specialized roles in the team. Teams become bigger and bigger. AI becomes an essential part of many, many businesses. You need to make these people to work very efficiently, right? I’m talking about data engineers, resources and data scientists, and ML engineers who are responsible more for automation as well as application engineers, software engineers, who is responsible for certain models. Those folks need to work together more efficiently, and they need a common set of tools. In many cases, those tools are based on GitHub and traditional software development tools. We just need to connect those tools with data tools, and we need to add some metrics-driven, data-driven capabilities for this tools set. This trend is clear because a lot of teams are struggling with collaboration. They badly need new collaboration tools.

Swapnil Bhartiya: Thanks for sharing these directions. Now if I ask you, what is going to be the focus of the company in 2022?

Dmitry Petrov: The focus of the company, it’s for us, data management to us. Focus number one from the beginning, in 2022, it is not changing. However, we will be focusing more on data-centric AI approach, which basically means, we will be focusing not only on data management, but also on label management. When people can build models by changing data, changing labels, and this way, do the modeling in opposite of changing architecture of ML models.

Data labeling is definitely focus for us and collaboration. Collaboration is a trend. People need more collaboration tool, not only among the data scientists, because there are a bunch of tools like how to collaborate among the data scientists, but we need to enable collaboration between engineers, data scientists, data engineers, and all the shareholders in data project needs to collaborate in one platform.

Swapnil Bhartiya: Excellent. Dmitry, thank you so much for sharing these predictions and also focus of the company in 2022. I’d love to have you back on the show next year to see how many of your predictions turn out to be true and get a set of new predictions for the next year. Thank you.

Dmitry Petrov: Thank you. Looking forward to hearing from you again in next year.

[/expander_maker]