Using Clean Rooms for Privacy-Centric Data Collaboration

Summary

• Privacy-safe collaboration: Databricks Clean Rooms enable secure data sharing across organizations without exposing raw data.
• Isolated computing: Organizations collaborate within governed, isolated environments with custom privacy policies.
• Compliance auditing: Clean Rooms provide complete audit logs of all activity, supporting compliance in regulated industries.

Using Clean Rooms for Privacy-Centric Data Collaboration

Talk by DJ Sharkey, Staff Software Engineer, Databricks ; Nikhil Gaekwad, Product Lead, Databricks
Databricks Clean Rooms make privacy-safe collaboration possible for data, analytics, and AI — across clouds and platforms. Built on Delta Sharing, Clean Rooms enable organizations to securely share and analyze data together in a governed, isolated environment — without ever exposing raw data. In this video, you’ll learn how to get started with Databricks Clean Rooms and unlock advanced use cases including: Cross-platform collaboration and joint analytics Training machine learning and AI models Enforcing custom privacy policies Analyzing unstructured data Incorporating proprietary libraries in Python and SQL notebooks Auditing clean room activity for compliance Whether you're a data scientist, engineer or data leader, this video will equip you to drive high-value collaboration while maintaining full control over data privacy and governance.
Here’s more to explore:
Open data sharing for data, analytics, and AI: https://www.databricks.com/product/delta-sharing
A new approach to data sharing: https://www.databricks.com/resources/ebook/a-new-approach-to-data-sharing
See all the product announcements from Data + AI Summit: https://www.databricks.com/events/dataaisummit-2025-announcements

Chapters

• 0:00 Introduction to clean rooms
• 2:52 Data sharing challenges
• 5:52 Clean room use cases
• 11:14 Demonstration in action
• 15:23 Privacy and security model
• 18:25 Key differentiators
• 22:13 Platform advantages
• 26:29 Customer stories and partners
• 29:15 Identity resolution solutions
• 32:06 Roadmap and conclusion

FAQs

What problem do Clean Rooms solve?

Organizations often need to collaborate on shared analytics and AI but can't share raw data due to privacy regulations or competitive sensitivity. Clean Rooms enable joint analytics in an isolated environment without either party seeing the other's raw data.

How do Clean Rooms prevent data exposure?

Clean Rooms execute queries within an isolated workspace that neither party controls fully. Results returned are aggregated or filtered; raw individual records remain protected. Both parties see only analytics results, not underlying data.

How do Databricks Clean Rooms enable privacy-centric collaboration?

Built on Delta Sharing, Databricks Clean Rooms allow organizations to collaborate across clouds and platforms without exposing raw data. Both parties run approved jobs on overlapping data in a centrally owned and managed environment, maintaining full control over what results are shared.

What use cases do Clean Rooms enable?

Organizations use Clean Rooms for cross-company analytics, joint customer insights, collaborative fraud detection, AI model training across partners, and identity resolution. All without sharing sensitive raw data.

What compliance benefits do Clean Rooms provide?

Clean Rooms address privacy regulations like GDPR and CCPA by providing comprehensive audit logging and custom privacy policy enforcement. Organizations can prove compliance through audit trails while maintaining full data governance in isolated, controlled environments.

Full transcript

0:00 Introduction to clean rooms

Hi everyone, welcome to this session. I hope uh you enjoyed the keynote. Uh there were a lot of exciting announcements. Um and I hope uh you guys get a chance to use the new products. In this session uh we have a very interesting talk about sharing data across organization. Uh so our team here, our speakers here are from datab bricks uh and DJ and they're going to talk about how to share uh data safely um and have privacy in mind.
So with that I'm going to hand it off to them. Uh thank you. Awesome. Thank you. Well, good morning everyone. Uh thank you so much for joining us in this session. I lead product management for data and AI collaboration. Uh joining me is DJ Shy, also engineering and tech lead for data and AI collaboration. Uh in this session, we're going to talk about how you can leverage clean rooms to be able to collaborate with your partners in a privacycentric way.
Well, one of the first things uh that comes to my mind when I think about data sharing is sort of the traditional data sharing methods like sharing files over SFTP or maybe you're using Excel and still sending that over email or if you're in the modern stack and you're maybe using something like delta sharing to be able to share data in place. Uh those are some of the methods of uh data sharing. But not all data can actually be shared across organizational boundaries.
Maybe it's like hypers sensitive or maybe that data has some regulations that you cannot share across your org boundaries. But of course there's a lot of things that you can actually unlock when it comes to data collaboration and sharing especially when you're sharing this across your customers or your partners. So the question really is is that is there a way for us to share data across organizational boundaries without revealing that data with our partners.
Is that even possible? Well, I'm going to give you the secret. Yes, it is through database cleaners. So in this session, we're going to essentially walk you through some of the challenges that organizations are seeing when it comes to data collaboration. Uh we'll give you an overview of our product which is data clean rooms and show you a quick demo how it works. Uh we'll also talk about some of the challenges when it comes to privacy and how do you use clean rooms to ensure your data is shared in a privacycentric way.
And last but not least, we'll talk about some of the successful customer stories that we've seen with clean rooms and also our partner ecosystem. So we have a pretty packed agenda uh to walk you through starting off with some stats. So when we talked to a lot of enterprises and this was a survey that was done last year over 93% of organizations said that it's really important for them to collaborate with data and essentially leads to higher levels of innovation and it drives revenue for a lot of uh enterprises but there's always a but a lot of enterprises also said that over 56% of them think that you know data sharing is difficult because of privacy and they have concerns about how they and make their data available to their partners especially when the data contains some sensitive information.

2:52 Data sharing challenges

So for this a lot of customers are again looking for alternatives to be able to collaborate with their partners while ensuring that data uh remains private. And what are some of the challenges that we're seeing across the industry when it comes to data sharing? Well, the first one is is that there's certainly some privacy and compliance laws that are continuing to change. So if you look at GDPR and CCPA or some of these regulations that are you know being enforced, it's really important for organizations to think about how do you ensure when it comes to data collaboration, you're remaining compliant and you're also respecting user privacy.
Again, you have a trust with your customers to make sure that the data stays private and of course it's only being used for uh the right reasons. And of course there's also risks when it comes to data security right with so many data breaches you have to ensure that you're you know storing your data in a centralized way where you can govern that data asset and most importantly staying in control of that data especially when it leaves your organizational boundaries is really important.
So data security is a big concern when it comes to uh collaboration for a lot of organizations. And lastly, complex operations and approvals. So, who here has gone through the legal process of essentially getting data across to a partner or signing some sort of data partnership agreement? Yeah, many of you. So, you can probably imagine all the red lines that you see uh with the legal teams when it comes to data collaboration.
So uh there's a lot of approval processes that need to be put in place especially when the data is leaving your boundaries and a lot of the traditional data sharing techniques I talked about SFTP and downloading the file in Excel and sending it over email again is not really uh compliant and it doesn't really help you scale especially when it comes to being able to innovate in real time and really driving uh value out of your data.
So what does this lead to? a lot of missed opportunities and certainly doesn't help you innovate uh within your organizations. So these are some of the challenges that enterprises are facing today and to essentially address this and make it easier for you to collaborate we've introduced what we call uh databicks clean rooms. So more slide uh the solution is clean rooms. Uh clean rooms essentially allows you to uh share data in a privacycentric way and clean rooms was actually introduced we actually previewed it in the keynote last year and as of today we are GA across all the three clouds.

5:52 Clean room use cases

So AWS, Azure and just two days ago in GCP so you can spin up a data bricks clean rooms across the three cloud platforms and invite and collaborate with partners uh regardless of which cloud or data platform they're on. So I'm gonna now let DJ Shy walk you through how data bricks clean rooms works and just give you a little bit of intro into uh how privacy also works within a clean room. DJ, take it away. All right, so did a great job of giving a quick overview of clean rooms, but you're probably all wondering what really is a clean room.
We get the idea. We get some problems we can solve. Let's go let's go a step deeper here. And you probably weren't asking what really is the architecture of a clean room, but I'm going to give you a quick overview of that. And then I promise we'll get a little more, you know, a little more use case driven here. So a clean room, you know, let two or more parties run approved jobs on overlapping data without actually sharing access to the underlying data or IP.
Behind the scenes, the whole thing's powered by delta sharing. You're delta sharing data from not to each other, but to this centrally data works owned and managed central environment. uh and delt sharing kind of an implementation detail but also really important because it powers some really cool stuff. It lets us be super crosscloud, super cross region, crossplatform even lets you use your clean room without moving your data, without replicating your data.
So let's go through an example use case here. This is kind of like your base use case within clean rooms. It's audience overlap between an advertiser and a publisher. Maybe some of you have had this idea, had this happen in the past, but you just want to know, hey, how many users do we have in common? So, an advertiser reaches out and says, "Publisher, how many users do we have in common?" Publisher says, "I would love to answer that for you.
Just send me your user list." Advertiser, understandably not so happy about that. I can't share my user list. It's my full footprint. It's, you know, sensitive information. There are a million reasons I I can't do this. can't and don't want to do this. Why don't you just send me your user list? Publisher that would solve the problem. Publisher also not so thrilled about this. So with the data clean room, you could actually answer this one specific question without sharing all of your raw overlapping data.
Basic idea is both parties add their data to the clean room. You add a notebook and you say, "Hey, can I run this specific query?" Um little picture of it on the screen. and you get an approval and then yes, you can answer that specific query and get that specific answer without sharing everything. Next use case here is building on top of just data protection. Now we're doing data and IP protection for identity resolution.
So here you've got an e-commerce company and an identity provider. E-commerce company wants to get some wants to work with the ID provider to get universal IDs. Maybe they're trying to create a customer 360. Maybe they're trying to use those universal IDs for like downstream activation use cases. You know, million reasons to work with an identity provider. Um, without a clean room, you go through this kind of song and dance of ID provider says, "Sure, I'd love to give you universal IDs.
Send me your PII." The e-commerce company says, "No, why don't you send me your full ID graph and your identity resolution algorithm, your proprietary ID resolution algorithm. ID provider of course says no to that. So what do they do? They use a clean room. And in this example, they both add their table. You add your PII table. Your ID provider adds their ID graph. And your ID provider also adds in what we call a private library which can be used to protect that underlying code um that underlying IP.
And now you can actually assign universal IDs without sharing access to any proprietary information or code. Uh we're actually also recently I think today announced uh some partnerships with some specific ID providers that Nikquil is going to go into later today or later in this very talk. Last use case here very similar to the last one but trying to drive home the idea that it really there are cleaner use cases in any industry in any vertical.
So this one we have a bank and a credit agency working together to identify fraudulent transactions. So, your bank has transaction data. Your credit agency has an ML model that can be used to rec use to recognize fraud. Your credit agency doesn't want to share the raw ML model. Maybe they, you know, they don't want to keep track of how often it's being used. Maybe they want to do something like or just make sure it's not actually like being copied.
They want they want to be in the loop. Uh, and your bank can't share raw transaction data. So with a clean room you could actually permit the usage of the ML model. So you could allow inference on the ML model without sharing direct access to the model. No risk of it being copied. And similarly without the bank sharing that transaction data. Awesome. So those are some detailed use cases. How does this thing actually work?
What does this actually look like? We're going to go into a demo now. And as with all great demos, it'll be the next slide. Give me a sec. As with all demos, we will start with a slide, right? Uh just to give you a quick over overview of what we're doing. We have a brand, a data provider, and a publisher all working together. Uh the brands trying to act trying to figure out, you know, the right segment for them to activate.

11:14 Demonstration in action

Uh you'll see you can see they're actually all in different clouds and Kil mentioned this, but Dicks Cleaner offers fully crosscloud capability, which means you can meet your collaborators where they are. No need to move data around. And with that, we'll actually start the demo. Okay, there we go. Cool. So, what we have here is a brand that's actually creating a clean room. There are three separate tabs there. It's going to go back and forth a bunch.
Apologies. I'll do my best to explain it. Give it a name. They add in the sharing identifiers, which are effectively crosscount IDs. They're actually metas IDs for those of you that really know your data bricks. create the clean room. Few seconds here it's created and it'll automatically appear for both the data provider who can now see the clean room and similarly for the publisher who can now see the clean room. Heading back over to publisher can now see the clean room.
Heading back over to the brand, we're now going to add data into the clean room. Brand's going to add their transaction level data. They know you know who purchased things from them. And really important call out here is when they add this data into the clean room, they don't share permission to query it. That's only done through these approved notebooks that we're going to show you later. Behind the scenes, this is Delta sharing the data to that clean room environment, not to the collaborators.
So, we're going to switch over to the data provider. Oh, we're in the data provider account. And we can see the schema. We have no way to query the data though, not yet. Same's true from the publishers point of view. All right, behind the scenes, other two parties adding their data. Not going to bore you with the same thing over and over again, but same idea. Data in the clean room doesn't mean you could actually see it yet.
So, we're in the brand now, and the brand's going to actually add a notebook to the clean room. This is kind of the magic of magic of clean rooms. that notebook has permission to query this data and you could eventually get it approve you could get it approved by your other collaborators and then you could run anything you'd like on it. All right, so now we're actually adding that notebook in. We're going to select ourselves and say this notebook is runnable by me.
Awesome. Let's see. Gonna actually add the notebook and I'm gonna go run the notebook. Yay. Not so fast. All collaborators must approve the notebook before it can be ran. This is as I mentioned before really intrinsic to the data bricks clean room security model. The idea is each collaborator has control over their data is is queried. So in a second here we will switch gears into the data provider's point of view. There we go into the data provider's point of view.
They're going to open up that notebook that they see it's been added to the clean room. They're going to look at its code and they're going to say awesome. This looks great. Approve it. And the publisher behind the scenes is going to do the exact same thing. See notebook has been approved. So brand can now actually run the clean can actually run the notebook and as soon as the job results in a few seconds here they could actually see all of the clean room compliant output. they could see the results of those queries that they that they ran.
And this is awesome. This is an example of a fully crosscloud clean room running approved queries on top of clean room data and only approved queries. Cool. So, we just saw a clean clean room in action. You might be asking yourselves, how does privacy in clean rooms actually work? And we'll go a little deeper into that. The key here is explicit approvals which allow for flexible use cases. I love the cleaner the data bricks cleaner approval model.

15:23 Privacy and security model

It can support literally anything at all you want. So here we have an example advertiser you know can write a notebook. Hey could I run this notebook this Python this SQL code in the clean room touches both of our data. Publisher could look at it and say yeah that looks great or no. Uh and once it's a fully approved you could run the notebook. What makes this so powerful is you have access to the full power of data serverless notebooks within your clean room.
If you have an idea, you can run it in the clean room. Super flexible, supports any industry, any vertical, any use case. Now, some of you might be asking yourself, wow, that sounds like a pain in the butt. What if I don't want to approve every single thing that happens in my clean room? What could we do for that? So we actually offer parameterized notebooks which will help you scale approvals. Basic idea here is just publisher and advertiser.
Again publisher maybe has some privacy enhancing technologies. uh maybe it's K anonymity, maybe it's aggregation thresholds, maybe it's differential privacy, whatever it is, they could enforce those privacy enhancing technologies in their clean room. They could use a runnerup supplied parameter and say, I'm only going to run this runnerup supplied SQL query if it passes my checks. So they get to totally define privacy.
And at the end of the day, you have one notebook, one approval used for many different flexible queries. You know, advertiser here is happy. They write and run a query. Step three, they write and run a different query with no additional appro approval. Fantastic. And then the last step here is restricted internet and output. Makes code review simple. Got a malicious company. They hand you a notebook. You're in a clean room in a clean room with a malicious company.
Don't do that. But you are. They hand you a notebook. They say you should run it. And you go and run it. Turns out this is also safe because your clean room actually run is a restricted environment with limited output limited places it could output data as a when you're running a notebook it's only sending output to you. This means it's actually safe to run things that as long as you're the runner, you can run it without risk of excfiltrating your own data and slide.
I'll take care of the slide. I think it's it's mine next. So it is. So at this point you're probably wondering, wow, I mean I hope you're wondering, wow, clean room clean rooms looks great. you're asking yourselves, well, what makes data bricks cleaners the best? And Nquille will help answer that question. All right, thank you. Uh, one thing again, just wanted to pretty much summarize what DJ just walked through when it comes to the clean room.

18:25 Key differentiators

So, uh, by definition, clean room essentially allows you to go in and analyze these disparate data sets that might belong to, you know, two or more parties in a privacycentric way. And what these just talked about is you know collaboration without actually sharing the underlying data. So in that demo you saw that none of the parties were actually revealing that data to the partner but you're able to run compute and run analysis on it and see the output.
So that's the real beauty of clean rooms. Uh the way that we control privacy and ensure privacy is through uh approvals. So we've built uh data bricks clean rooms in a more flexible manner which means that you can essentially control the privacy based on the notebooks that are made available ensuring that all the parties in that clean room are approving that notebook before it's run. So our privacy is essentially focused on the approval of code and this really gives you the flexibility especially if you're you know doing a lot of complex use cases right because we support SQL and Python uh you can essentially orchestrate any use case that you can think about in clean rooms and we'll see some of the use cases that we've seen uh in the industry so far and then last but not least it gives you that extra layer of enhanced security.
So uh when you actually create a clean room in data bricks behind the scenes we actually go create a new account metas store that's managed uh by data bicks for that collaboration which means that any data that's made available in that clean room only the notebook has access to it and it's able to run the compute. So none of the parties that you're collaborating with can directly see uh or access the clean room. So uh it's really important for you to take this away because this is what sort of clean room unlocks uh especially when it comes to data collaboration.
Uh what are some of the use cases that we've seen for clean room? So uh we actually seen clean rooms get adopt across the various industries. Uh the early adopters of course have been in the martekch and adtech industry. Think about the deprecation of device identifiers and uh maybe the deprecation of thirdparty cookies which has gone back and forth. Uh you know we've seen many of these industries lean into clean rooms as a way to be able to uh make their first party data available to their partners and run analysis uh without again ensuring that it's it's privacy safe.
So if you look at advertising and retail you have like audience segmentation. So if you're planning a a joint campaign with another partner, you can share the first party data from both the partners and plan your your audience that you're going to activate on uh measurement and attribution. So once you run the campaign, how do you go in and measure the effectiveness of the campaign itself? So uh clean rooms essentially helps you unlock many of those use cases.
In financial services, um actually we got a shout out at the keynote for Mastercard. We've seen that they're starting to leverage clean rooms to be able to again collaborate with their partners and data providers uh to be able to do things like fraud detection or make sure that they're meeting the regulatory and compliance uh that they have for being able to collaborate with data uh healthcare and life science. I know probably some of you are here from that industry.
You can certainly think about you know PHI data that is hyper sensitive uh that you cannot collaborate or share with your partners. uh again clean room provides you the mechanism to be able to still unlock the rich insights from multiple partners uh ensuring that you know you're you're remaining compliant and of course in manufacturing supply chain there's again a ton of use cases that you can think about uh when it comes to clean rooms so DJ essentially started off by saying you know what makes data works clean rooms unique and we actually started off by thinking about you know what are some of the pain points customers have for organizations ions have when it comes to existing clean room solutions.

22:13 Platform advantages

Again, transparently, we're not the first clean room solution in the market, but we have an advantage because we can actually learn from uh some of the other clean room solutions how they've been deployed and essentially build uh the capabilities that help resolve some of those common pain points. The first one being, you know, not all your partners are going to be in the same data cloud or your data platform that you're working on.
Some of them again if you're on AWS some of them might be in Azure or GCP or they might be using Bitquery or Snowflake. So it really you know becomes difficult for you to collaborate with them especially when you have to replicate the data and oftentimes you're locked into the vendor uh that you're working with and going to your partner and saying hey you want to work with me but you have to replicate your data to the cloud that I'm on is a difficult conversation.
It also doesn't provide the flexibility that you need especially when you think about some of the AI use cases right so um a lot of the clean room solutions today are SQL first so you can only run you know restricted SQL on the data but again if you're thinking about some of uh the AI use cases especially if you're training models or doing inferencing uh you want to be able to use more flexible languages and existing clean room solutions don't provide that today and also as you start to think about how do you scale this across across uh not only hundreds but thousands of partners that you're working with again it might not might be that first clean room that you have with your partner but as you're thinking about okay how do I do this across all the partners and customers that I work with uh you have to think about scale and automation and some of the clean rooms today are uh restricted in terms of how they enable you to scale so what we did at data bricks is we wanted to go in and address all these challenges and pain points head-on so the first thing to note is we essentially allow you to collaborate across any cloud or any data platform.
So you can spin up a data bricks clean room today and invite your partner regardless of which cloud and platform they're on and they can still participate without having to move or copy their data. Yes, they would still need a data bricks account because again they're governing their data asset. So they would still need a database account to be able to essentially connect to that data. But once it's made available in Unity catalog, they can collaborate in a clean room environment again without needing any replication.
Second one is because we're essentially using notebooks, you can use any language that you wish. So we today support SQL and Python, but you can we're also going to be supporting uh JVM and JAR files very soon. So again, think about those complex orchestration or workflows. Those will be available uh directly within notebooks and within data clean rooms. And last but not least, we've sort of built clean rooms from APIs ground up.
So if you want to do this at scale or maybe you want to build your own UI, which is part of your whole data collaboration platform, or if you're a retailer, this is part of your retail media network of how you deliver insights with your suppliers and CPGs. So clean rooms essentially allows you to take our APIs, remain compliant uh with essentially sharing your sensitive data but again have your own experience that you want to build on top of it and of course if you want to scale this across you know hundreds of partners you can use built-in orchestration workflows.
So big differentiators for data bricks again we've addressed many of the common pain points that customers have had for clean rooms head on. uh one important again thing to focus on is this ability to collaborate again across clouds and across platforms. I think this is the really big pain point for a lot of customers um and organizations that I speak to is again if I'm working with someone that is not in the same cloud how do I do that effectively so what we actually launched is what we call lakehouse federation where you can essentially connect to the data in place and make it available in data bricks clean rooms and still still collaborate and DJ essentially showcased that uh in his demo so big takeaway here so talking about some of the partner uh and ecosystem and customer stories.

26:29 Customer stories and partners

So, as I mentioned, we actually previewed clean rooms last year. Uh since then, we've gone G earlier this year and we actually launched GA for uh GCP a couple days ago. Uh we've actually had a lot of customers that have leaned in and chosen data bricks as their core uh clean room platform. Uh starting with uh in it. So in it also has a session here where they're essentially leveraging data clean rooms to be able to uh work with uh many of their partners.
Um some of their challenges that they had was a lot of the external vendors that they work with you know could not access the sensitive data that in it collects. So uh and they wanted their partners to essentially use what they call marketplace 2. 0 know uh where they can look at some of the prospects of various customers and provide back some leads that those partners can can take and work on. So, in it has essentially built uh capabilities in data with clean rooms where their partners can essentially train models on sample data, bring that model into a clean room and then essentially run inferencing on into its first party data again without into it revealing or making that data available uh to these partners.
So, uh really successful I think uh shout out to honor and team that have worked really uh hard here. There's actually a talk about this as well for in it that if you're interested on how they did it, you should certainly check that out. And of course, as I mentioned, there was a uh keynote u shout out from Mastercard as well. So, Mastercard again has chosen database clean rooms to be able to uh again think about this as the core way of how they get data from external vendors uh into their platform and how they ensure that whenever that data is made available internally or externally this sort of becomes a core way that they can uh go in and facilitate sort of data collaboration.
So there's a lot of use cases that Mastercard is thinking about when it comes to clean rooms especially when it comes to monetizing data products. So uh again they've done a lot of great work here and many of the things that Mastercard has built is of course what we call privacy enhancing technologies on top of data bricks cleaners. So if you think about how do you ensure that data the hypers sensitive data remains private uh you can use things like K anonymity right it's a technique where you're anonymizing data certain threshold or differential privacy where you're injecting noise into that data uh Mastercard has done a lot of work to essentially build this in as part of the notebooks and again surface that to their partners when it comes to cleaner so a lot of great innovation there and one of the more sort of mature clean room use cases that we're starting to see is uh what we call identity resolution where some of you might be aware of this.

29:15 Identity resolution solutions

This is essentially where you know you're essentially have first-party data like PII data uh about your customers and if you want to think about you know building out a customer 360 view of you know all the different data assets that you collect or you're thinking about perhaps you know anonymizing the PII data for you know further you know use cases like audience segmentation or activation or personalization identity resolutions becomes really important And traditionally what customers have done is they've essentially taken uh their PII and sent it to the identity resolution provider via SFTP and essentially trusted the ID provider to be able to take that data resolve it and send back their corresponding IDs.
Again, not a great way to collaborate, right? Because essentially sharing data that is hypers sensitive, which is your PII with the identity partner. Again, you're trusting that partner with that data itself. So, uh what we've started to see is many organizations say, you know, can we do this in place? Can we essentially run identity resolution in place where uh the identity resolution provider can read our PII and convert that into the corresponding ID which we can then use for things like activation, segmentation, etc.
Uh so legacy pipelines you know uh are are not great and lead to a lot of inefficiencies. So what we've done is we've actually worked with many of these large uh identity resolution providers uh including epsilon, axium, live uh and deote where they're actually bringing their technology within database clean rooms where again the ID provider brings their identity graph and their proprietary code for example and they can make that available within a data bricks cleaner. you as a customer can bring in your PII data uh into the clean room and just run one single notebook.
Um running that notebook would essentially read your PII data, resolve that with the corresponding ids and then you can see the output. And the beauty of this is that again you're not moving or copying the data out. It stays in place if you already have your data in data bricks. You're not again replicating this uh in a different environment. And of course you're can do this at scale. So especially if you're running this on a regular basis as your customers and graphs get updated, it becomes really efficient for you to continue with this process.
So we're super excited about this particular capability and use case unlocking uh a lot of u innovation for customers when it comes to cleaners. So just to kind of wrap up before we get into Q&A, uh I just wanted to say that, you know, we're certainly investing a lot in the clan room space. So as you could probably imagine, we announced preview last year. We've already gotten to G and we launched things like multicolorator uh earlier this week and uh you'll see many of the announcements that are coming up.

32:06 Roadmap and conclusion

So, lots of innovations and clean rooms, especially when it comes to being able to scale, when it comes to bringing sort of data privacy. Uh, and I've labeled out some of the things that are on our road map. Uh, one of the things that we're excited about is we're actually looking at, you know, integrating uh, some of the clean room capabilities into databicks marketplace. So again, if you are a uh partner or a customer that's thinking about, you know, how do I get my partner to be able to access this clean room and start uh working with me efficiently, you can imagine they can go to a marketplace, find your listing and start the clean room collaboration with you and you can sort of facilitate many of the process.
Especially if you're an ID provider, again, this becomes a lot more efficient for you to get customers on board and start giving them access to again uh your IP and services. So really excited about that. Uh talked about u output. So DJ essentially demonstrated how we have the ability to output u the result set of a clean room for the runner itself. But we're looking at providing more flexibility. So maybe you might have use cases where the output of a clean room gets delivered to two or more parties or both the partners at the same time.
So uh some of these capabilities again we're hearing from customers about uh these use cases and we started to invest and build uh many of these capabilities within data bricks. All right with that uh want to conclude and say thank you so much for listening and this is the end of the talk. I'm happy to take any questions you guys have.

Learn more about the Databricks Data and AI platform.