#264 Will GenAI and Data Mesh Really Mix? - Interview w/ Madhav Srinath

Please Rate and Review us on your podcast app of choice!Get involved with Data Mesh Understanding's free community roundtables and introductions: https://landing.datameshunderstanding.com/If you want to be a guest or give feedback (suggestions for topics, comments, etc.), please see hereEpisode list and links to all available episode transcripts here.Provided as a free resource by Data Mesh Understanding. Get in touch with Scott on LinkedIn if you want to chat data mesh.Transcript for this episode (link) provided by Starburst. See their Data Mesh Summit recordings here and their great data mesh resource center here. You can download their Data Mesh for Dummies e-book (info gated) here.Madhav's LinkedIn: https://www.linkedin.com/in/madhavsrinath/In this episode, Scott interviewed Madhav Srinath, CEO at Nexusleap.Overall, we are super early in the Generative AI cycle and hype is huge. This discussion is one of early impressions, not fully formed answers. It's far too early for that.Also, FYI, there were some technical difficulties in this episode where the recording kept shutting down and had to be restarted. So thanks to Madhav for sticking through and hopefully it isn't too noticeable. Generative AI will mostly be shortened to GenAI throughout these notes. LLM stands for large language models which power GenAI.Some key takeaways/thoughts from Madhav's point of view:?Controversial?: An emerging best practice seems to be having layers of LLMs - one model where you might ask it complicated questions and the second model is trained specifically to vet the answers for correctness and governance concerns.The cost of running many models in production is typically actually quite low, at least infrastructure wise. Instead of an always-on architecture, most organizations are leveraging a serverless architecture - or leverage APIs from others providing the models - so they essentially only pay a few cents per query.?Controversial?: Use GenAI as a "scalpel, not a broadsword". Many are trying to use them in overly broad ways and getting not great results.The ability to take a mountain of data and get something out of it in a structured way isn't a new concept. We've been trying to do that with data mining for years. It's just that it is finally...