Boosting Data Science and AI Productivity With Databricks Notebooks
Summary
- Collaboration Hub for Teams: Databricks Notebooks serve as a central workspace where code, visualizations, documentation, and results coexist, enabling faster iteration and development.
- Multi-Language Support and Integration: Interactive coding features, Git integration, MLflow tracking, and command shortcuts accelerate data science workflows and experiment management.
- Real-Time Collaboration: Version control history, commenting, sharing, and co-editing capabilities empower teams to collaborate efficiently and deliver AI projects faster.
Boosting Data Science and AI Productivity With Databricks Notebooks

Talk by Jason Cui, Senior Product Manager, Databricks ; Vijay Raghavan, Senior Director, Thumbtack
Want to accelerate your team's data science workflow? this video reveals how Databricks Notebooks can transform your productivity through an optimized environment designed specifically for data science and AI work. Discover how notebooks serve as a central collaboration hub where code, visualizations, documentation and results coexist seamlessly, enabling faster iteration and development. Key takeaways: Leveraging interactive coding features including multi-language support, command-mode shortcuts and magic commands Implementing version control best practices through Git integration and notebook revision history Maximizing collaboration through commenting, sharing and real-time co-editing capabilities Streamlining ML workflows with built-in MLflow tracking and experiment management You'll leave with practical techniques to enhance your notebook-based workflow and deliver AI projects faster with higher-quality results.
Databricks Named a Leader in the 2025 Gartner® Magic Quadrant™ for Data Science and Machine Learning Platforms: https://www.databricks.com/blog/databricks-named-leader-2025-gartner-magic-quadrant-data-science-and-machine-learning
Build and deploy quality AI agent systems: https://www.databricks.com/product/artificial-intelligence
See all the product announcements from Data + AI Summit: https://www.databricks.com/events/dataaisummit-2025-announcements
Chapters
0:00Introduction and overview2:52Notebook features demo12:33Thumbtack use case and security20:56Securing notebook workflows
FAQs
How do Databricks Notebooks boost data science productivity?
Databricks Notebooks create an optimized collaboration environment where code, visualizations, documentation, and results coexist seamlessly. This unified workspace reduces context switching and enables faster iteration, allowing data science teams to deliver projects more quickly.
What collaboration features are built into Databricks Notebooks?
Databricks Notebooks include real-time co-editing, commenting capabilities, revision history, and Git integration for version control. These features enable teams to work together efficiently while maintaining code quality and project history.
How does MLflow integration help with machine learning workflows?
Built-in MLflow tracking and experiment management within notebooks enable data scientists to automatically log model parameters, metrics, and artifacts. This integration streamlines ML workflows, improves reproducibility, and makes it easier to compare and track model iterations.
What multi-language features does the platform support?
Databricks Notebooks support multiple programming languages including Python, SQL, Scala, and R, allowing teams to use their preferred tools. Command-mode shortcuts and magic commands enhance productivity across all supported languages.
Full transcript
0:00 Introduction and overview
All right, good morning everyone. How's it going? Cool. Thank you for coming. First session of the day. So, funny thing about this session is they told us this was supposed to be the only occurrence. It got completely filled. Uh so, thank you for bringing that energy. So, we're actually doing it again tomorrow if you don't get enough today. So, we're here to talk about notebooks. Who needs coffee? if you have notebooks.
Um to kick us off, I'll just walk through uh some logistical things. Here's a quick forward-looking statement uh around this presentation. Um as well as Oh, we were supposed to plug the app, but this slide isn't there. So, uh there we go. Uh before we start, a quick intro on me. I'm a product manager here uh at Data Bricks. I have the pleasure of working on our notebooks product. uh we actually have a lot of our team here uh who has worked on uh the amazing features as part of this product.
So if you see any anyone with like a white employee badge, be sure to say hi. Um give them your feedback and we'll all wrap back to our team. Uh but I have the pleasure uh to be joined by Dr. Vay from Thumbtac. Um he I'll sort of be passing the mic over to him midway through uh to talk about Thumbtac and some of their use cases. But uh yeah, welcome. Uh thanks again for coming so early in the morning. Uh the title of this talk is boosting data science and AI productivity with data bricks notebooks.
And so maybe quick show of hands. Who here has used a data bricks notebook once in the past week? Okay, pretty good. Uh that's actually more than I expected. So you all are familiar with the product and maybe I don't even need to give this whole talk. But uh I think we'll sort of walk through some of the fun new features we've been working on and then also talk about how we think of notebooks strategically uh within data work.
So you know what is a notebook? I think you know even for me and my own data science journey the first time I touched a notebook was back in undergrad when I took an intro to data science class and uh you know I fell in love with Jupyter notebooks but I think the ecosystem has really come a long way since then and I think the way we think of notebooks at data bricks is more of like a window into our data intelligence platform and the reason we think that is that first notebooks are the most used product at data bricks just by a pure user count right like folks are in notebooks every day.
They're doing various things from data analysis to building pipelines. So, it's really an extensible and powerful platform for folks that are within our ecosystem. Um, I think beyond that, our notebooks have evolved again from the Jupyter notebook to become quite powerful and collaborative. And so, we'll walk through some of the specific features there around how notebooks have really been enhancing in terms of um functionality.
And so before I get into the live demo um of the specifics, here's a quick slide that we walk through um around why notebooks are simple um and powerful. So first I think with our data bricks notebook, we really try and focus on a modern intuitive UX. Those that have been data bricks OGs will know that we actually did a whole refresh of the notebook um making things a little bit brighter and easier to use. Um the second is that the notebook is actually infused with intelligence.
2:52 Notebook features demo
So folks here who are familiar with assistant um that is deeply embedded into our notebook and we will have a separate session uh covering functionality around the assistant itself. Um and the final thing is we're extending beyond just a simple python notebook or SQL notebook and we really want to become this sort of embedded IDE experience within uh data bricks bringing some of those powerful features um that you might expect there.
So, I know that's a lot, but you're probably wondering like what is new in notebooks, right? Like maybe some of you see uh our blog posts or or see like our what's new section within the product. Uh I'll cover some key pillars today, but then of course if you have any questions after um this talk today, feel free to come up and chat and we can talk through specifics. So, first uh we'll talk through a little bit around serverless and our strategy there uh specifically within notebooks. uh we'll talk about some features around an improved developer experience uh within notebooks and then finally we'll briefly touch on um sharing and collaboration.
So okay enough slides uh I'll quickly switch this here and it should throw up my demo. Perfect. Can you all see that there? Awesome. All right. So first serverless right I I think um we've been launching we've been working on serverless for over a year now and I think our overall vision here is that hey at the end of the day like who wants to configure clusters right it's it's a painful experience it's difficult um the vision with serless is to make compute simple fast efficient and reliable within the notebook and so to give you an example uh I have a sample query here and uh let's hope this works but you'll see I'll run the query um I wasn't previously attached to any compute Uh the whole goal with serverless is that it should automatically spin up and start uh quickly.
So we'll see here. So you'll see I'm already connected. That was you know a few seconds uh to kind of spin up and connect to serverless. And then hopefully if the connection here is working uh this should uh this query should run. So the last time I tried this out uh the Wi-Fi cut and uh I think the demo gods were not very kind to me. But uh okay, it seems like this is actually running. So let's give this a few seconds.
Great. So now you have this beautiful car data set. Um which if anyone is looking for a new car, I can share this with you afterwards. Um but yeah, again the whole purpose with serverless is to put compute into the background, right? So for folks who are not really attuned to switching between clusters, um this this will be readily available in the notebook for them and they don't have to do any um thinking there. Um, folks who are newer to the notebook, uh, I'll point out some specific features, uh, in like the paneling here.
Um, and so you'll see I pulled in this car data set. If I go on the left side here and go to my catalog, uh, that data set will show up as active. So really great if I just need to reference it. Um, if I need to insert the table name uh, into a particular cell, uh, I can actually open this table in the catalog explorer itself if I want to um, just look at it or see any sample data there. So little bit of a preview into serverless and um the catalog functionality there.
Um to talk a little bit more around some of my favorite features actually. So we talked a little bit about flexibility, right? So right here I have a SQL cell um but I'm sort of more of a a Python lover myself. So, one thing that I I will frequently do is like, you know, I'll pull in this table. Uh, I'll get a quick look at the data, but then I'm like, okay, I what I really want is to have this in a Python data frame to be able to play around with it.
Uh, maybe transform it. And so, you'll see I have a Python cell right below this SQL cell. Um, and I mentioned right the assistant is sort of embedded directly into the notebook. uh I can just quickly uh use a keyboard shortcut to sort of invoke the inline assistant and I can say actually like okay it seems like this table is stored in this sort of SQL data frame. Uh let's just say like we want to access that um and display the head and the assistant should generate the right code for that.
Uh I'll just accept that and run it. Cool. So I have this data available now in a Python data frame. So this is a pretty common workflow we'll see where folks can sort of reach for the languages and the workflows that they're they're used to. So if you're used to pandas, if you're used to polars, all of that is accessible uh within the notebook for you to use. Um and uh it sort of intermingles together. All right, talking a little bit about developer experience.
So, like I mentioned, we did a big UI overhaul of the notebook uh in the past year, but we our team has also been working hard at basically pursuing these feature enhancements to make your developer experience in the notebook much easier and much more powerful. So, the first thing you'll notice that I actually didn't call out um is that right now in my notebook experience in the workspace, I actually have tabs. So, this is available right now uh in production.
It is default off. Um, so if you want it on, you'll have to go and explicitly toggle it on. Uh, but the main benefit of this experience is rather than going through the painful experience of having tons of browser tabs, uh, I can actually open all of my data bricks assets within the workspace. And you'll see here I can switch between tabs, between notebooks, uh, between files and just making it more of an IDE like experience um, that folks have used are used to.
So, this was one of our um most commonly requested features and we're very excited to bring it um to you all. Um I think one thing I personally love about tabs is right now like if I were to go and go into search anytime I search for a file or open it uh it'll show up in the tabbed experience just making it a lot better uh a lot easier to actually look through those sorts of files. The next thing that I'll highlight on the developer experience side is sidebyside editing.
So, this is sort of fresh uh off the press. It's still in a private preview, but you'll see if I take one of my tabs and drag it uh to this other uh region, it'll open my files side by side. Um so, again, I mentioned our team that worked on this is here. So, if you see them, you know, give them give them some kudos. Uh but this is, you know, again, fresh off the press here. Um, again, one of our really really uh commonly requested features, but you can imagine uh this enables like a whole suite new suite of powerful workflows.
Maybe you have one notebook open that's your main working notebook. You can have a separate reference notebook open uh where you can look through helper functions or things like that. And so um I had a ton of fun uh playing around with this feature. Um we're we're constantly enhancing it and we'll introduce like an icon entry point there as well. Uh so then uh you can go ahead and you know split your screens there. Uh the final thing I'll call out on the developer experience side is command pallet.
And so you'll see that we have an icon here uh on the bottom right. If I click that uh it'll actually invoke the command pallet. You can actually do it via the uh keyboard shortcut as well. Uh but I'm just calling this out because this is one of my favorite new features as well. So you can use it both for navigation um to menu settings. For instance, if uh you're like, "Hey, I want to uh create a new view or something and I forgot where to go in the menu to do that."
Uh you can use it to change themes. Basically, uh anything that's accessible in the menu is accessible there um in the command pallet. So, uh again, uh just bringing some improved developer uh experience there. The final thing I'll call out is sharing and collaboration. So I think the reason sharing is so important to us um at the notebooks team is like we view the notebook as like a living breathing document right it's a data artifact for you to maybe start some analyses but you know the whole goal is to reach some sort of insight and so the target flow there is to go do your analyses and share your notebook uh with your team and so one feature I wanted to call out is we completely revamped our comment experience within the notebook and so you'll see here like if I go into any particular particular code snippet.
Um, it can be code. It can actually be markdown as well. Um, if I go and comment, uh, I'll say this is a great point. Uh, we've completely revamped uh, the UX here. Um, to be like a Google Doc sort of experience uh, where folks can go in, uh, they can reply to comments and you can actually tag folks in here um, and you'll be notified via email uh, just as you would um, through your Google Sacks experience. So yeah, just wanted to call that out as as a final point um of this demo.
So I know that was quick. I could probably talk about notebooks all day. Uh but I wanted to leave enough time uh for to walk through his content. So maybe before I'll transition it off, I'll ask uh any favorite notebook uh features from you uh in recent recent usage. Uh pretty much I love them all. Um the one thing I found to be very very useful is how responsive the data bricks team has been. I constantly bother uh with ton of request and pretty much they've been able to indulge us with almost pretty much all of our apps.
So thank you for that. Great. Okay, that concludes the notebook demo. Uh I'll hand it over to uh to walk through the next portion of this. Excellent. Thank you. Excellent. Uh, how many of you here um use Thumbt Tag or heard about Thumbt Tag? Cool. So, Thumbtag is a marketplace for services. We bring together home professionals and then um homeowners and our job is to make sure that any issue that you have with your home, we understand it and then our professionals get matched to the right type of a job.
12:33 Thumbtack use case and security
So, pretty much that's it's a it's a Uber uh for your home services industry. Okay. So, our our mission is um helping the homeowners and then home professionals accomplish more overall. So, at the core of it, we are an AI company. Um pretty much everything if you look at our app and then if you look at our website, if you look at our matching technology, pricing, AI is part of everything that we do almost on a session basis.
So one of the things that powers all this AI infrastructure is data bricks notebooks. So uh we use pretty much everything. So we use uh a couple of models from GCP. We use all the foundational models open AI cloud whatever and pretty much the connecting in bedrock is one of the major uh part of our infrastructure workflow as well. Pretty much the connecting theme across everything is our data bricks uh notebook workflow. uh all our multimodal search today runs out of data bricks notebooks.
All our multimodal uh customer experience in production also runs out of data bricks as well. So at the end of the day um we use ton of u LLM uh type of models today. I got a I I want to ask three quick questions to everybody. What do you think is a cost of a simple data bridge? Uh today let's say you have a data breach. Uh what what will be an approximate average cost of the data br uh data uh breach today? Any guesses?
Again, your employees won't be identified. So completely safe. I'm sorry. She's right on the money. Yeah, it's about 4. 88. Um IBM did a study last year and then for last of whole of last year, the average cost of a data bridge uh was about 4. 88. 88 million. So, excellent. Thank you. How many of those data breaches do you think or let me put the question in a different way? How much how many you know how many data breaches um typically are can be attributable to an issue with your AI model or some kind of a data leak from an from your AI model.
Again, open question. Um no judgments here. Feel free. 5% 2%. Let's say there are 100 data breaches. How much of that is attributable to a model or data leaking from your model or something bad going on at on the data side from your model? I'm sorry. You almost killed me, man. Uh so it's about 25% today, but I expect that number uh to kind of increase. So it's about 25% today almost you know um for about every uh 100 data breaches that you have uh especially on sensitive PII customer information about 25% of it is directly attributable to a model uh that's from a Verizon study that they did last year as well.
Question number three, compared to compared to 2022 to 2024, can somebody tell me what is the percentage of increase in these type of breaches uh that is attributable to a data leakage from your notebooks? 20% is not actually not a bad number. It's very close. It's about 300%. So Mckenzie did a study in 2024 uh a couple of actually they released a couple of months ago uh it's about attributable number of data breaches coming off that is directly attributable to a notebook uh kind of like you know data leak is about 300% or something uh very close to uh 300%.
So the common thread is everybody all of us think that I will spin up a notebook I will do something everything is going to be for experimental purposes but in reality what happens is everything that we do even for experimental purposes gets into production at some point in some way shape or form. So what I'm going to talk about today is how do you kind of um how do you kind of go about securing your end-to-end workflow when the whole workflow depends on some type of notebook.
So there are about you know uh eight in total uh that I can think of uh feel free to add more but uh the number one thing that you could do to secure your whole notebook infrastructure is make sure you centralize your LLM workflow. So everybody here um I mean how many of you know work with an LLM or some type of a multimodal model uh in your notebook or through your notebook environment on a daily basis. Okay, that's a good number.
And how much how much of that work involves some type of sensitive data in the notebook itself or pulling some kind of sensitive data through a notebook? Just raise of hands. Nobody. Okay. So these are the brave souls. I I I bet 70 to 80% of you uh use some type of sensitive data data in your notebook. So the first thing that could be done is have a good security layer. So you centralize all your workload into one cluster one set of uh one set of way way of doing things so that you reduce the attack surface.
Number two is uh a very simple solution but often overlooked solution is to make sure that your secure workplace is underneath um isolated separately from rest of your uh rest of your infrastructure notebook infrastructure uh through a VPC security u layer number three could be data set permission. uh for example one of the one of the easiest way to do it is to make sure that you do it explicitly not implicitly. So one of the solutions that I've seen that a lot of good enterprises do today is to make sure that like you know they manage data as they manage any other part of their infrastructure as well.
So one of the common things among good organizations that you'll see today is using Terraform to manage um uh manage data permission data set permission at a very granular level. uh if you want that. Number four is using some uh some type of scrubber. Even if you are not planning to use PII level data, even if you're not planning to use anything sensitive, I would say for any of your notebook infrastructure, notebook type of work, um I would highly recommend um you using some type of scrubber so that even by accident, you don't leave uh something that can be exploited later.
And of course observability has come a long way. Um at thumbtack for example everything that we do uh there's an observability framework that is built upon our notebook infrastructure from data bricks and also uh through fiddler. So everything that you need that you do on a daily basis make sure that like you know you have some kind of observability set up for that. Okay cool. So um in total um I can think of like I said eight uh but just for demo purposes I'm going to demo four of them.
20:56 Securing notebook workflows
So in total the eight um eight things that you could do to make sure your notebooks are secure. Like I said I'm going to go over the list and then I'll show you the demo uh in a little bit. First, centralize your LLM workloads, LLM infrastructure uh on your data bricks workflow. Uh number two, make sure that you will have uh some set of firewall rules uh set up. You know, putting putting it under a VPC uh is actually a good first step.
And then there's other things that you could do. Um number three is make sure that your data set permissions are set up in a way that for people who need access to X, they have only access to X, not Y and Z. Um and then the third thing is always make sure that you have some kind of a security scan um setup for your libraries that you import and whatnot. Uh number five is the PII scrubber uh that I just spoke about. I'll give you a demo of a PI scrubber, a simple PI scrubber that you can kind of set up.
Uh number six is also PII monitoring. It is not enough to set up a scrubber. it is not enough to set up something and then you know uh not to observe or monitor uh your models for data uh leaks as you kind of um go and then move things into production. So setting up some kind of a PI monitoring setup is fairly useful. Uh seven is you know um I think it's one of the uh best practices that is often suggested by data bricks uh which is your uh setting up your secrets manager.
Uh in fact like you know a long time ago um I was bothering who's our data data bricks amazing data bricks u account executive that we work with on setting up a learning plan for my team and the first lesson that you know usually that is offered through a learning plan uh is how do you manage your secrets in your notebook so it is very very important that you don't avoid this first step I would say it's worth the investment in terms of um doing it and number eight is tried and uh tested key rotation u policy.
So I've set up an instance. So you know um everything in the back end is taken care of. So here the demo I'm showing is if you set up everything right and then if your VPC setup is working um you people cannot just go ahead and then get data or give data to an external URL. So here what you will see is I have I'm trying to connect to two instances. One is something that is uh tried and tested, approved by my team and approved through various security measures and then the other is not.
Uh for example, one is whitelisted uh which is the API from OpenAI uh where the connection was successful and then the other was an request to uh just a GitHub repo of some sorts. And obviously you can see that the GitHub repo request times out and then the OpenAI uh stuff is working. I cannot I cannot um stress this enough. Anything that you do remember you one of the common misconceptions is I write this notebook today and then you know this is just for uh P kind of purposes.
I just want to test something really quickly and nobody is ever going to use this notebook for anything else. That's the most common misconception because what what happens typically is many people will see that notebook and then they will take that as a best practice. They might not think twice about copying that notebook to something else and before you know this notebook will have a life of its own and then it would have been forked n times over.
So one of the common things that I do um apart from my regular job I also um I also research actively on security of AI models and I also uh teach classes to mostly people from our government agencies uh through a partnership with George Washington University in DC and one of the first thing when I give when I give um u exercise to my students is even though let me remember these people work full-time for our security agency And then I just observe how many of them can will um use a library to scan for something and then scan for PII before submitting the assignment.
So the answer is not that heartwarming. But anyhow, I digress. Uh coming back to the point, I cannot stress enough of how much of a uh how much of a good thing in general and how much of a trouble that it will save you. uh how much of trouble uh is say how how much u uh trouble later on will be saved by using something like you know a very common uh PII scrubber so what you're you know uh pretty much like I would say 70 to 80% of any of our personalization models most of the models that you use today require some kind of personalization would have some type of an identifying information for example um in the case of thumb tag And often times uh we see uh people talking about their addresses uh people talking about their home.
So these are personal. We want to make sure that these things don't get out and these things are managed effectively and I'm pretty sure every this is a common theme across each and every of your org. So one of the best practices is to make sure that like you know you have a PII scrubber and then actually um you have very simple mechanisms to do that today. Uh, one of the things I would recommend is an open source project from Microsoft called Precedio that you can use.
Of course, we had to uh we had to do some modifications for a few thumbtack use cases specifically, but um the way we kind of do it is um we took the open source model uh from Precidio. Uh we did some changes to reflect of uh one of our couple of our business use cases. to make sure that it is done the right way uh per thumbtack and uh we use an artifactory uh manager. So we publish it to artifactory and then that's how we kind of download things uh because you know downloading any libraries out in the open is an issue as well.
God knows um who touched that library before you. So here what I've done is basically um I'm installing I'm still I'm installing all the libraries that I need and then what you will see here is so I'm using uh you know u I'm using my own internal bucket to download everything and then what you will see here is uh a bunch of text that has some type of um some type of personal information for example somebody's talking about the credit card number uh somebody's identifying the place that they live.
Uh somebody is uh somebody is talking about an email that they sent from an IP address. Uh somebody's talking about much more identif identifiable information, passport numbers and such. Uh somebody's even talking about their uh bank account number for payment details. So don't don't underestimate how much people can how much people are trusting of your business and then willingly give information. uh there is people people trust businesses a lot the underlying uh trust behind that is we know that you will handle my information in the right way I don't have to worry about it so in this case look how I have kind of scrubbed all that information out using one simple library everybody can do that it's one of the standard best best practices.
Uh but in reality, I don't see it getting used that much. So I would highly recommend this as a best practice for all your notebooks. So this is all the information that I sent. I use my scrubber. So the scrubber pretty much removes everything. And then there you go. But it doesn't stop here. you always need to worry about what happens in the background and really uh you need to have a robust way to observe these things in the real world.
So like I said before uh our whole um ML/AIN infra uh runs through data brickres uh but we do have an observability platform that we work with fid we work with a company called fiddler a lot uh in terms of making sure the observability part of everything that we do is taken care of. So there is a there is a there is a good set of monitoring mechanisms that you have today. And if you look at fiddler the orange and then the blue lines.
The orange one is probable PII information uh that is sent over in my notebooks or present in my notebooks uh before I use the scrubber. And the blue line tells me how much of that is has been removed after scrubbing. So this is very important because you need to understand the fundamental nature of the data that you're kind of handling. So I would recommend for any of your models that is going into production or even if you're doing something that is you know um something that is us and throw kind of a deal when you're kind of um analyzing certain data sets or even simply uh looking at things like data drift in your data or whatever.
I would I would highly suggest that you monitor the heck out of anything that even resembles PII because this these are the things simple things that you can do today that will save you lot of heartburn 6 months from now a year from now. So of course uh secrets manager uh is something that is um highly recommended and um like I said we we are a very um hybrid environment. We use GPC, we use uh bedrock, we use open AI, we use cloud.
There is lots of things that we do but everything that we do uh there is no um nothing that is kind of floating out in the open. Everything is managed through uh a secret uh secrets manager. So overall, eight things. First, reduce your attack surface. Uh centralize your LM multimodal workloads as much as possible. Number two, make sure that your VPC/ firewalls are set up so that you cannot just your egress um how people get and load libraries and all that um is kind of controlled in some way, shape or form.
And third is make sure that you're using simple things like terraform to make sure that you set up your uh data set permission um as much as controlled as possible. And then four is make sure that like you know whenever you use a library it's a common best practice uh is to have something like you know artifactory kind of a workflow to make sure that people don't use libraries willy-nilly and then they have uh the right type of li that you can kind of load and many organizations um have also started started auditing their library use as well.
So this will be useful for that purposes. And number four um you know uh sorry number five is the PII scrubber that I was talking about before. Uh simple thing like precedio can be very very helpful there but there is also lots of other uh tooling u for your notebooks and observe everything. Don't take things for granted. I would say this is where most of the people don't I would say uh it's not so much about not doing the right thing. it is uh I need to ship something tomorrow so I you know I'll do something in a notebook and then I will kind of get it over with but uh these are the things that will kind of uh bite you and monitoring in general is a good practice um even for your data drifts and things like that some type of observability monitoring mechanism um is is recommended uh seven is secrets manager um I cannot recommend that enough number eight is key ro key rotation so pretty much to sum up everything if you do this eight things or even do 50% of these eight things in your normal notebooks workflow as a best practice before you ship everything.
Um you know pretty much you will be uh very secure in your not u notebook work kind of workflow. So that's it from me.
Learn more about the Databricks Data and AI platform.
The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.