Databricks' Youssef Mrini: Democratizing Data + AI

0:00Hi everyone, we here at Data Plus AI Summit and look who are with me Yousef, my very good friend. Yousef, welcome to the Rawat show. Such a pleasure. Great announcements at the keynote. But uh we're going to get more into it. Uh but before getting into that, would you like to tell us a little about what you do at Databricks? Yeah, for sure. But first of all, I'm super happy to be here with you. I've been following for quite some

0:22time like watching all your recordings and they're just amazing. Thank you. Thank you for all the work you're doing for the data community. you've been like I've been following for quite some time and it's just amazing. Thank you. Thank you first for this amazing uh amazing job. And uh yeah talk a little bit about me. I'm Yousef. I'm a solution architect uh working for for Databricks and I'm

0:44based in Paris and I cover mainly startups and DNB. So if you have any question just let me know. I'll be happy to help you with all of your issues. And Yousef really is uh the most approachable person when it comes to you know uh talking about Databricks uh the different problems helping the community and much more. So you can count on that. Uh yousef um great announcement this

1:07morning at the keynote and uh I I'm super pumped about listening all the great announcement talk about the free edition talk about you know agent brakes talk about uh mlflow 3.0 a lot many more things that were announced. Uh I would love to know a little bit from you as well. What do you think about the keynote announcements and uh tell us a little bit about the Databricks free

1:29edition? What is that? Yeah, so basically Databricks revamped uh the community edition. Nice. Like we're created a new version called free edition where anyone can just go spin up a workspace within few seconds. It can literally fit in a real or a short that's super super fast and they can access all almost all of Databricks

1:52features for free to make sure that they can practice or increase their knowledge. So this is like an easy way to get started with Databricks. Yeah, that's fantastic. And it also kind of gives a lot more opportunity to those who are wanting to do make mistakes and learn from it, right? Uh so it becomes an interesting platform. Uh so congratulations on that announcement. It

2:15definitely helps the larger community out there uh to get their hands on Databricks. So I'm also curious to learn a little bit about Databricks AIBI and also the Databricks one. Uh we should uh jump right into it and um my you know first thing that I wanted to ask is around uh AIBI and Databricks one are about democratization and getting data

2:37and AI into the hands of all users including the business users. What do you think about it? So first let's talk about AIBI. Uh it was announced one year ago and since then like a lot of features came in. First we rebranded the dashboard. Now it's called AIBI dashboard. Why? We're bringing AI to the dashboard to make sure that every single person can build

3:02dashboards easily. Y basically you can just text hey I don't know uh do a chart of like revenue by product and you're good to go. no longer need to have this deep experience in dashboarding to build a cool one. Yep. And the second part is about Genie. Like Genie is like the way where we can go further and discuss with your with your data. It's like literally

3:26talking with a data analyst who can write SQL queries for you and that's it like hey I want to know like how many sales I did last year. You have the answer quickly. Yeah exactly. I think sometimes you know uh going and reading through the dashboards kind of also become a challenge and if there's a genie that can really help you to get through and get your answers really quickly why not uh right yeah yeah and

3:49that that was that was the purpose of my talk yesterday was like saying dashboards are good they can help you with so many things but what if the question you're looking for is not answered by dashboard right so Genie will cover this part and you can just ask a question and you will have fast answers And we have been releasing many many features uh to to improve the

4:12AIB genie like we're talking about something called knowledge store knowledge store right yeah where you can just like bring talk about synony deploy your synonyms uh define your joints uh comments it's like basically a way to empower you to increase the um the um the availability or the AC the accuracy of your general right I love it uh you

4:34those were fantastic points and I totally agree with that Also, I'm kind of curious because you talk to a lot of customers and I'm pretty sure you get a lot of feedback. You also uh understand what's working for them, what's not working. Uh what have you been uh observing are like the biggest challenges to being you know truly datadriven across entire company.

4:57Can you share some um something along those lines? Yeah. Yeah. So, this is something I kept saying every time. Datab bricks has been built by the community for the community. Right. Right. We keep hearing feedback feedback too. And those feedback are transformed into actionable insights that leads to

5:19build features and you mentioned something earlier about uh Databricks uh one one. Yeah. So datab bricks one is the answer to datab bricks being simple and being able to be used by business users. Yep. Because now if you go and open you're like a business user want to use Databricks sometimes can be overwhelming with all the innovations we

5:41have inside you're going to find like I don't know workflows compute orchestration geni good yeah it's a good problem to have to honest I was happy to have this problem and with Databricks one it's lightweight version of Databricks where you just deploy or give them access to only what they need and for instance it can be for example access aib genie like because you can

6:04deploy it with APIs and maybe have Genie within I don't know Teams or Slack. But what if you want to grant them access to the Databricks one? They can just log in and they can use Genie or maybe access some dashboards, right? And it's it's super easy and it's accessible to any to anyone. Yeah, exactly. I think uh that's a very important point when it's kind of accessible to everyone but it's also easy to use and Exactly. I love how

6:27you kind of also mentioned about the community. It's for the community and it's always been for the community, right? Uh so that's fantastic. Uh also what success trends have you seen with uh companies that double down on democratizing uh analytics and AI? Any thoughts around that? I think the Databricks like did a lot

6:49of uh improvement to the platform to make sure that they democratize access to data to every single person. I think years ago we started with data engineers. Yep. then like data data scientists then we shifted again to data analyst and now we are targeted targeting business users and basically we can use the platform for your entire

7:12data journey from the ingestion y till the till the deployment and this can also help the customer because once they have like a unique platform it's much easier to integrate all those cool components and get value out of them. Yeah, I think uh that's very important uh when you kind of getting the value for the customers and for the business users. It is something that they kind of

7:34you know really are looking out for and uh database is kind of helping them to uh you know reach the goal. Uh in the last 12 months how is Genie kind of evolved? Any uh thoughts? I know you mentioned a little bit about it but do you want to get into the features? Uh anything specific that you've been seeing that the customers are most excited about? I think it's hard for me to keep track

7:57all the updates and the releases they've been uh like releasing for the couple like last couple 12 months. I remember I was talking with the uh with the marketing team. they were like literally struggling to keep up with all the updates like and this is again this is a very very good problem like as I mentioned I mentioned the knowledge store where you can define your joints comments synonyms they release something

8:20called benchmarks like this is basically those SQL queries you can define like you ask you you you ask a question right with the query and you can test your genie like quickly to make sure that it works well for you they released something called like history where you can share your feedback what where whether it's a good response from Genie or maybe a bad response and if it's the case how we can discuss with person who

8:42triggered this this this this issue they release the APIs like to to move it to production and having this possibility to integrate it with Slack teams or what or maybe datab bricks app if you want so I think the like the innovation is just like it is really an a good opportunity to work for datab bricks and to use Databricks because you're never bored that is something for

9:05sure yeah uh this fantastic I think uh also I want to switch gears a little bit around the Databricks one as well. I know for a fact where uh it kind of is positioned for a lot of marketers, sales teams, a lot of uh other teams in the space who are very datadriven too. How was uh it a challenge before this for them? Do you want to shed like a little

9:27bit of light on that? Yeah. So basically in the past we had two options either grant them the access to the workspace and it is it was still I mean a good option but because like your data is governed with unity catalog so you know that this specific group of people won't be able to access maybe the sensitive data or maybe data from

9:49another team but it was cumbersome for them because of all the innovation we have. Exactly. It was a little bit hard for them. So they had two options either use the Databricks workspace or they can use uh I don't know teams or slack but sometimes they want maybe to have another option and one Databricks one

10:11is the option yeah I think uh that kind of solves a lot of problem for them as well and if they can now uh you know they can be more datadriven freely as well about things and the work that they've been doing in the space uh those are fantastic insights Coming to a little bit towards the governance piece as well. Uh there there are also a few updates around the unity catalog for

10:34unified data plus AI governance. What do you think? Um uh and why do you actually think the companies should be prioritizing unified governance right now? Uh what are your thoughts? Yeah. So again if you go if you come back a little bit in the past you had to use many tools like uh a tool for data warehousing a tool for ML for ETL for

10:58dashboarding and this led to discrepancies because you had to copy the data multiple times to handle different uh governance systems which lead to like errors. Yep. with UC or with the data intelligence platform. We did combine all of those uh all of those components in one and that's what we

11:20call the data intelligence and the awesome facts about data intelligence is everything is governed with unity catalog which means you define the the governance once for your table your like not only your tables but your AI asset as well right and then you can use them across all the services and it's much more easier and more secure so not only you know who can who like have the

11:43permission to access data but also when which is super important when it comes to security. Yeah. And how how how do you think the companies should approach uh the you know the AI governance uh in the AI era? Yeah. I think it's it's super critical right now because in because we have a lot of data sometimes can be sensitive data and you don't you don't want to have like

12:05someone from another business unit maybe accessing this data. So that's why you need to make sure that you're granting the right permissions to the right people and this is very crucial and that's why Unity catalog is super important because it's very easy to use like you can do it with the UI with the API like depending your like your knowledge or how advanced you are and then it's again used for all with all

12:29the components. Yeah, that's fantastic. Any favorite features from the Unity catalog that you would like to mention? I know there are so many. Yeah, just a few maybe maybe two or three. Yeah, I think the first one is uh the Aback the attribute based access controls where basically you can define now uh tags. So I can for example go to a specific

12:51column and say this column contain PII. So I will just tag it PII and I can tag another one maybe uh uh sensitive data or whatever and then I will just define a function say hey whenever I have PII data please mask it or hash it right which means no longer need to duplicate data for the same table you can have multiple multiple uh like return results

13:15because of those policies you would have set up I love it uh one last question for you and thanks for sharing all the great insights you're always as good as uh you know when it comes to data and AI you're a top class so thanks for sharing all the great insights one last thing uh if folks want to reach out to you learn more about the different things that you're doing I know you share a lot of content you have a podcast you do a lot

13:37of different things uh and you work on the products as well and you work with uh clients too so a lot of things uh where can they reach out so first of all uh reach out to your Databricks representative if you have if you are a Databricks customer if it's not the case feel free Sure, feel free to reach out to me in LinkedIn, Twitter or the YouTube. So my name is Yousef Marini as

13:59Ravidit mentioned and the YouTube channel is NextGenLakehouse. Fantastic. I love it to answer all those questions. Thank you. Appreciate it. And uh like I said, Yousef is super approachable, very cool guy. So go and follow him and uh love all the content that you put out and thanks for sharing all the great insight. Such a pleasure hosting you. Thank you so much. Thank you. Thank you everyone for joining us.

14:21Bye. All right.