Large-scale machine learning and AI on multi-modal data

The Data Exchange Podcast: Bob Friday on incorporating machine learning and AI into large-scale, real-time applications.


Subscribe: AppleAndroidSpotifyStitcherGoogleAntennaPodRSS.

This week’s guest is Bob Friday1, VP and CTO at Mist Systems a Juniper Company.  Bob is a serial entrepreneur and seasoned technologist, and at Mist  his team uses data technologies, machine learning , and AI to “optimize user experiences and simplify operations across the wireless access, wired access, and SD-WAN domains”. This means having foundational data infrastructure in place to ingest, process, store, and analyze hundreds of terabytes of data daily.

Download the 2021 NLP Survey Report and learn how companies are using and implementing natural language technologies.

Bob and his team build models from structured, semi-structured, and unstructured data. They have deployed anomaly detection models that rely on deep learning (LSTMs) and have begun exploring the use of graph neural networks for a variety of use cases. They have also built and deployed systems that use recent advances in natural language models. Their virtual assistant provides insight and guidance to IT staff via a natural language conversational interface.

Bob Friday:

Machine Learning has been around for a long time, we’ve solved a lot of problems. I think the big transition is “are you really doing something on par with  humans”? … We’re basically trying to take domain expertise, and stick it into Marvis. We’re trying to build something that actually does something on par with a network domain expert.

… I’m a big fan of what they call human-assisted AI. …  you’re talking about an AI assistant, where the human domain expert still solves the corner cases. You use AI to solve 80% of the problem, but there’s still 20% of the problem that requires some domain expert. And to some extent, that’s what we’re doing. Marvis solves about 70% of the problem;  30% of the time we still need a domain expert to come in and help push it over the finish line.

Download a complete transcript of this episode by filling out the form below:

Highlights from the video version:

Related content:


[1] This post and episode are part of a collaboration between Gradient Flow and Juniper. See our statement of editorial independence.

[Image by Oleg Gamulinskiy from Pixabay.]