Sponsored by: Dataiku | Agility Meets Governance: How Morgan Stanley Scales ML in a Regulated World
Summary
- Agile governance balance: Morgan Stanley broke the false choice between compliance and speed—combining Dataiku and Databricks for rapid, governed ML.
- 50% faster insights: No-code workflows empower business users while scalable infrastructure handles terabyte-scale ML with full audit readiness.
- Cross-functional success: Unified governance and risk automation enable data leaders, ML practitioners, and compliance teams to innovate together without friction.
Sponsored by: Dataiku | Agility Meets Governance: How Morgan Stanley Scales ML in a Regulated World

Talk by Raja Lanka, Executive Director Cloud Tech, Morgan Stanley (Sponsor Speaker)
In regulated industries like finance, agility can't come at the cost of compliance. Morgan Stanley found the answer in combining Dataiku and Databricks to create a governed, collaborative ecosystem for machine learning and predictive analytics. this video explores how the firm accelerated model development and decision-making, reducing time-to-insight by 50% while maintaining full audit readiness. Learn how no-code workflows empowered business users, while scalable infrastructure powered Terabyte-scale ML. Discover best practices for unified data governance, risk automation, and cross-functional collaboration that unlock innovation without compromising security. Ideal for data leaders and ML practitioners in regulated industries looking to harmonize speed, control, and value.
Here’s more to explore:
Unified and open governance for data and AI: https://www.databricks.com/product/unity-catalog
See all the product announcements from Data + AI Summit: https://www.databricks.com/events/dataaisummit-2025-announcements
FAQs
How did Morgan Stanley achieve both agility and governance?
By combining Dataiku's no-code workflows with Databricks' scalable infrastructure and governance, Morgan Stanley reduced time-to-insight by 50% while maintaining full audit readiness for regulatory requirements.
What are the benefits of no-code workflows in regulated industries?
No-code tools empower business users to develop and deploy ML models without IT bottlenecks, while maintaining governance through centralized data, model, and audit tracking.
How does Databricks support terabyte-scale ML?
Databricks provides distributed computing infrastructure to handle massive datasets, Unity Catalog for governed access, and MLflow for model lifecycle management at enterprise scale.
Who leads this innovation at Morgan Stanley?
Raja Lanka (Morgan Stanley Executive Director, Cloud Tech) discusses how the firm's wealth management division transformed ML development through the Dataiku-Databricks partnership.
Full transcript
Okay, let's get started. Hi, good afternoon everyone. Thank you very much uh for making the special effort to make the journey and uh show up in the room. It's really really appreciated. Thank you. And uh I really hope to make your journey worthwhile here and uh share some insights into how we have uh solved a very big business problem that most of us have in our day-to-day tech industry. Right. So before we get started, it's again a pleasure to be here in San Francisco.
A gateway to Pacific is what it is called. And uh I know this place is now teaming with a lot of uh tech enthusiasts. The maximum number of tech enthusiasts are in and around this area. So it's a great place to be and uh exchange ideas, learn new things. Always a pleasure. Thank you very much again for the uh pleasure of your company here. So let's show with get with a quick show of hands. How many of you have felt like technology, compliance and innovation are acting as blockers rather than enabling each other.
Anyone? Okay, that's that's a quite quite a good number of hands. So that exactly is the loop we have actually broken uh in Morgan Stanley. We have converted that loop. We broke that loop and converted it into an express way where we can resolve issues uh for our business using technology ensuring 100% compliant solutions. I'll uh take you on a journey with me on uh into Morgan Stanley wealth management where we have enabled a unified platform using the power combo of data bricks and data IQ and how we have enabled a huge set of users to be able to extend their boundaries and deliver solutions.
So before I get started, I am an executive director at Morgan Stanley with uh 14 plus years of experience at Morgan. I've been uh in various different office locations in Morgan Stanley and I'm currently based out of Atlanta office where we have acquired Erade and we have formed a significant presence in Alpharetta region of Atlanta. I know data analytics, machine learning, all of these could be a very serious topics.
Okay, the topic itself is serious but we can have some fun while we are learning about it and I'll make sure that we keep this session lively and given this small audience, we can even make it interactive if you all wanted to. For those of you that are not familiar with, I represent Morgan Stanley, a global leader in financial services with a legacy spanning nearly a century. So since 1935, Morgan Stanley has been operating in investment banking, wealth management and institutional securities.
And we have always strived to be on the cutting edge of technologies and uh with a commitment to innovation and excellence. For us, delivering best-in-class solutions for our customers is of utmost importance. And the way we do it is we always stay ahead of our ahead of the curve and innovate solutions. And just to give you an example or a metric about it is we were the first financial firm to launch a machine learning model into production that has given us the next best action for our customers.
How do we what do we say to the customers? How do we interact? That has been uh decided through a machine learning model that has given us the top five ideas we can do uh we can discuss when we call call up our customer. You can uh read more about it if you Google genome and Morgan Stanley you can read more about it. So that's uh that's led to a individual-based marketing from a segment based marketing and we were the first ones uh out of all the financial firms to deliver it.
But at the same time we also understand that uh in our mission to give a best-in-class service we also have to make sure that in today's world technology and the financial uh finance are have become inseparable. They are joined at the hip and they cannot go in either direction. Right? They both have to have the same kind of journey. So before I uh dive into the uh solution, I just want to establish what are some of the core challenges we have been facing and how we have uh identified those and resolve those.
So some of these challenges I'm sure you have heard about them in the last 24 48 hours and you have experienced yourselves in your day-to-day life in your whatever firm and whatever domain you might be coming through like uh especially something like a talent management or uh the uh market competition etc. and the overall implementation costs on how to keep the costs in check. Say for example you you cannot add hardware and unlimited hardware to solve your big data problems.
As we all know the data has been growing and the data growth has been really significant in the past few years and not just the in terms of volume but also in terms of variety the growth has been significant. So one of the biggest use cases we have seen is in our day-to-day life we have seen that our structured data use cases are outnumbered by our unstructured data where we want we need capabilities to be able to handle log files uh call call logs images uh PDF documents and all types of documents.
So our unstructured data volume growth has been so significant and the ability to handle such an unstructured data sometimes requires special compute like GPUs. We have uh use cases where we need GPUs to be able to run some of these use cases and uh that is a bigger challenge because creating a new GPU cluster or adding GPU resources on prem has always been an challenge in our data centers. So these are some of the things as as we discussed you might all be facing and this and the second one is like the talent management.
I just want to touch upon this where there there used to be a day where people used to learn one programming language C or C++ and they would have uh survived their entire career. But nowadays every second day the technology is evolving. People are uh people are learning new things. we there are new technologies that require us uh to update our skills. So it's an extremely important thing that uh our business people that understand the business be able to solve the business problems or create solutions that does not require them to have any specific technical skill.
So that's a big challenge we have been seeing. So in all of this the uh as we discussed one of the things is the innovation velocity time to market. We might be having the best of the ideas but if we are not on time into the market with your solution we are definitely going to lose out. So it is extremely important that it's not only having a good idea but able to take it all the way from ideation into production and into the customer's hands is the time is uh what we are calling it as time to market and keep it in a checked fashion.
So that is where I think innovation velocity is extremely important for us. But while we are innovating, we cannot overlook the fact that compliance is being followed. We cannot uh because we are in a regulated industry and we want our data to be safe. We want our solutions to be scalable, secure and all of it. So it's extremely important that we also take care of the compliance. So we can we don't have the luxury of choosing one over the other.
They both have to go hand in glove. So how did we do this? So this is where I think we are going to talk about uh the next the next problem that we have is uh the three-headed dragon is what we are calling it. So one is silos. While we are trying to solve this on this journey we also identified that most of the our teams are working in silos there. Some of them are comfortable using Jupyter notebooks. There is the other set that is working only on Excel and then there is a third set that is only dealing with documents.
So similarly like they they are they are comfortable with their own tool set and they don't want to uh explore other options. They want to be limited in their tool set. So how do we take them out of that silos? How do we give them the capability to be able to do much more than just operate on Excel sheets or just operate on Jupyter notebooks is a bigger challenge. we try to solve and we discussed about the compliance and governance.
So sometimes when whenever there is a new model development or a new solution that we are trying to develop uh how many of you have dealt with MRM model risk management in a traditional machine learning sense I'm sure you all have heard about it or dealt with it in your day-to-day life. So that's a big challenge. So the time taken for from ideiation to productionizing the model is sometimes less than the time taken to get your model reviewed through the MRM and getting the approvals because MRM is a resource inensive and a time consuming process where every document uh everything in the model has to have explanability. it has to be documented and it there were multiple scenarios that have to be run before the MRM gives us the approvals.
So our intention was never to cut short the MRM uh scrutiny but at the same time achieve the same MRM scrutiny within a shorter duration. And how do we do that is something we have uh started working with uh when we created a unified platform using data IU and data bricks. So we created one platform for our users and allowed them the flexibility or freedom to choose on which platform they want to operate and how do they want to operate with some standard guidelines that we have already set up for the users.
And one of the biggest uh outcomes that we have achieved out of this unifi unified platform is the semi-automatic documentation of data IQ has helped us to reduce the overall MRM process and it was a we were able to uh bring down the time time taken for the MRM review. I'm going to give you specific examples in the next few slides. But this is the unified platform I was talking about. The biggest use case here is we have given the flexibility for our users to be able to make that choice whether it is data IQ which is a low code no code environment.
People does who do not know or understand programming languages or do not have the required skills they are still able to get onto the platform and create complex workflows create complex recipes that is what we call it as that is nothing but a solution on data IQ and also create some modeling so whereas there was a time when modeling was always talking spoken in uh some secret rooms saying you need data science background to be able to model but we have democratized that model that is what is uh the democratization we have achieved.
I know you have uh if you have attended the keynote uh that is one of the most spoken word today democratization but we have been able to achieve that in real sense by creating a platform that allows people to create complex models without writing a single line of code. But at the same time it also helped us to keep track or keep the audit trails of what they are doing because that is how we have set up the platform marrying it with the data IQ capabilities and our own internal tools within Morgan Stanley that track the entitlements provisioning we call it as EPR enter entitlements provisioning requesting system.
So it has its own audit trail. Everything is logged. If a user uh raises a request to get access to a particular attribute that is logged and audited according to their entitlement provisioning. I'll also talk about arbback which is the role based access control. So that's that's another thing we have achieved but I'll just give you a highlevel view. So with these two platforms what we were able to achieve was we have achieved two levels of access at a very high level.
One is Aback attribute-based access control and the other is arbback role based access control. So what does attribute based access control mean? So at Morgan Stanley we are firm believers of having least access possible and in any given environment we encourage people to go for the least access that is possible or required. So our attribute-based access control helps us to go at the granularity of a column level. It is not even at a table level.
So if there are two PII columns social security number and date of birth and a user comes in and they only need access to date of birth they will only get access to date of birth and the social security number is dynamically masked. So that is the attribute based access control that we have achieved and similarly based on the roles people can access data and that is what we are calling it as role based access control.
We are going to talk a little more about it in the future slides. Second, uh just want to touch upon the databicks platform that we have created which helped us solve the biggest problem as we discussed challenge is scaling out the compute. scaling out special requirements or special compute that needs. So for example, we were we are running a use case where we need to do signature verification, document matching and all which needs GPUs and all of those special use cases used to take weeks if not months to execute in our environment.
But with data bricks we are able to create those clusters on the fly use them for the for the duration that we need and then shut it down. So that has helped us change the entire uh game in this uh solutioning and the overall architecture. So I'll also talk about uh the unity catalog and other uh aspects of it in the future slide. So I'm just uh uh I'll move on to the next slide. So this is uh the architecture we built.
Uh I'm going to show you the architecture diagram. But before that I just wanted to give you an high view of how we are enabling our users to be able to choose for business business reasons they can start using data IQ where they can uh get access to production grade data. So that's another aspect of it. So as we discussed we have created uh a sandbox environment which is one of our use cases I'm going to discuss which has a readonly access on the production data based on the user entitlements.
If users have the entitlements they will be able to see the uh confidential and PII columns and if they do not every uh all those columns will be masked and they'll be able to see the rest of the data. So that is how the data eq interface helped us and we were able to move the same solution onto data bricks as well with the help of unity catalog and all we are able to get the masking. Uh we are currently working on the implementation of it but we have already done the pilot of running the masking on dynamic masking on data bricks as well.
So our users then just make one request and they are autoprovisioned on all our platforms. So because it is they do not have to go and select each individual platform. So it it is a seamless access if the data element is accessible for an user based on their entitlement they can come through and access it on data IQ or they can go through data bricks and also access the same data element. So moving forward I uh would like to show you the architecture.
Uh this is the architecture for our advanced data analytics and machine learning use cases. So let's start in the left. We broadly classified our user community uh whether it is business, tech or anybody into five different categories. At the beginning of the spectrum is the data analysts who could be a fresher joining the team in maybe Budapest or maybe in Shanghai, Hong Kong. They do not have an understanding of our domain data, domain knowledge or domain data or any of that but they still have the need to create a dashboard.
They still have to create a report for uh maybe uh the senior management or they might even have to run a stress testing scenario which is required based on the uh requirements that are shared by fed. So that is how our data analysts operate. So and on the far right corner if you talk about data scientists they are the PhDs of data science. They have uh full understanding of data science. They want to be able to write their own complex models the like two to two to models DLRMS and advanced algorithms that will predict sometime do a predictions and sometimes even prescribe uh the next set of actions and stuff like that.
So all of these users are our user form our user community but as you can see here 80% of our users are classified into no code. They may code a little but most of them 80% of them fall under the no code umbrella and uh these people still now even in today's world they have the capability though they cannot write the code they can do any of these actions as you can see here they can do a data prep they can do data uh model development they can even deploy models and all of it that is because of the flexibility offered by data IQ platform.
The same thing goes again with the uh the other spectrum of the users like data scientists, the ML engineers and the data engineers who might have created hundreds and thousands of uh pipelines in during their life lifetime. they also they have the ability uh flexibility to come on board and start using data bricks where they can write the code on the pl the language they choose whether it is R, Scala, Python, SQL any any language they choose. uh we we give them the flexibility to bring in the libraries that are scanned through the X-ray and they will be able to use those libraries and they they can uh create complex models as you can see uh at the at a functionality level whether you're coming through data IQ or through data bricks your fun your core functions remain the same you are able to do the same activities by writing your own code or writing uh no code at So this is the architecture we have put up but now we stretch this a little more further.
Uh as I said I'll be talking about unity catalog and data storage and we have all kinds of special computes and all but all this special compute is only available on data bricks. So what if a data analyst creates maybe a very enthusiastic data analyst creates a model that requires GPU that's a big question right so that is also a problem we solved by using what we call it as DB connect so that's a special connector from data bricks that help that helps us to run any workflow recipe or a model that we generated on data IQ platform we can push that down into datab bricks and use datab bricks as an autonomous compute so it's just an compute layer so you do not have to write the code on datab bricks you can still write the code on data IQ and use the DB connect to push the data to push the processing of that particular workflow or model onto data bricks and it executes using the special GPU comp uh clusters of data bricks.
So that's that's a next big thing that we have done which has helped us to maintain this platform in a very flexible manner. So it it has helped us again just to give you a high level number it has helped us launch close to uh 40 models using this architecture in a very short span of uh 12 to 18 months. So moving on from the architecture I would like to discuss a few use cases as uh as we have uh as we are going through this.
The first one I would like to show you is the sandbox use case. As we discussed, sandbox or playground we call it both the names. It is an environment where we create which is created to allow our business users to access production data. So if any data that is on prod if they are if they need to run any report or if they need to create any model on that they can access it through the sandbox. The way we have implemented it is it is platform agnostic.
It's a concept for us. It is not specifi specifically linked to any one platform here. We're not just talking about say for example Hadoop, Terod data, snowflake. So whether you are on cloud, whether you are onrem. So some of these some of these are on-prem like say for example our Hadoop and Terod data still reside on prem and our snowflake is on cloud. But irrespective of that the sandbox architecture is spread across all our data stores.
Whether it is Hadoop or Snowflake, we enable them. We create views on the production data that needs to be exposed and grant access read access on the views for our sandbox users. Now they can take that production data and start creating their own versions. And just to give you a number to put it to put it in context, they create close to 100,000 tables per week for every 5 days they create 100,000 tables. We have we have a huge set of business users that are using this sandbox.
That is around 1,500 to 2,000 business users are using our sandbox environment actively and on a weekly basis they create 100,000 tables that they keep reiterating. They change one parameter here, one stress scenario there and they keep recreating the data in multiple iterations. They transform, they curate, they create it until they are happy with the output. So we have the our own retention policies and all and u on any given day our sandbox data that these people create is almost equal to the size of our production data.
So because sometimes they create these data and we need to retain them. So we have again that's a different uh discussion for a different day but we have a beautiful retention system and all where people can go and uh self-define retention policies on the sandbox data and if they don't define any retention policies we just drop the data because it it is it can be recreated by rerunning the workflow. So the uh second point I would like to touch upon here is how how we have uh implemented sandbox is another story where we have created 250 projects we call them PRJS.
So we have created 250 projects. So all these 2,000 business users that we are allowing access to them, they come in with their business use case or if there is a new requirement and if we we have a discussion architecture review discussion and if we deem we create a new project for them or repurpose an existing project and these projects they all work in their own swim lanes. There is no cross-pollination. Yes, they can show share the data but sharing of the data or collaboration of the code has to happen with approvals but otherwise everybody is in their own swim lane they start using uh the environments as I said these projects are created across all the platforms that are available whether it is data IQ data bras hadoop snowflake all these platforms have these projects and they're able to run their stress test scenarios they're able to run their fed regulatory reporting they're able to run their machine learning models all of those using these uh projects.
Uh and we we did discuss about the attribute- based access control, role based access control and how we are maintaining the integrity by only granting them readonly access. All of that is what is followed through the audit trail. Everything is tracked and logged for any auditing purposes. Moving on, uh we are talking about the use case number two. This is an interesting use case about a risk model that required I I'll start with this number here.
As you can see, this model used to run for weeks together. The people used to run these models. They used to go home and come back and check on Monday. They used to run it on a Friday, go home and check, come back next week and check if the model actually executed or not. And most more often than not, they used to have failures. Uh these are some investment uh office models and all that we have global investment office and other things.
So these were really really resource intensive and uh they had they created the biggest hardware machine available physical machine available on our in our in all of our data centers. Even then the model used to take days. So what we have done is we have worked with the business. We have taken the model rewritten uh some of the code using dev gena and other aspects of it and we were able to transition that model into this unified platform and create autoscaling clusters using data bricks and parameterized workflows using data IQ and we were able to bring down the overall execution of the model from weeks into less than 30 minutes.
That is because of mainly two things. One, when they were running these models on prem, they were not using the distributed compute capabilities. They did not have access to them. They did not have access to that kind of resources. And since we were able to bring them onto the platforms where they had access to distributed compute and they were able to create use special computes like compute intensive or memory intensive uh clusters and all we were able to bring down the overall execution time and now the model runs in less than 30 minutes.
Moving on uh the third use case I would like to discuss is the stale reports. Uh this this is another interesting use case where our business requires us to create uh the users to create a lot of reporting and dashboards. The underlying data was becoming stale for them because they were not able to uh as you can see they were not able to run these reports on terabytes of data. So that was a big uh blocker for the business and we had to we resolved it again by enabling them on we moved the dashboarding onto data IQ and we moved the uh SCA run terabyte scale run ML running onto training onto data bricks.
So now the as I said the dashboard resides on data IQ but the underlying training actually happens on data bricks for realtime data. So that's a big uh jump for us and now there is no more stale reporting and all the reports are uh run on realtime basis and in interactive mode. If they need to change something uh some parameters and pass additional features they do that and rerun those reports reports on the realtime data.
In the interest of time moving moving forward to the fourth use case. So this is the model we discussed uh I briefly touched upon where there are documentational needs. There are uh MRM uh requirements for every model and uh the way we have created data IQ role based access and attribute based access. Some of these have been standardized in such a way that while the model is being built we started generating documentation.
We started creating the documentation on based on what is the model as what how the access has been granted how the entitlements are being governed and how the PII elements are being masked so that the model does not use any PI elements in case if the model does not use or if it uses it'll also tell you what are the PI elements the model is using. uh so that is how we are able to uh bring down the overall documentational uh lead time and be able to generate models that are ready to be audited once the model development is done.
So this is another use case which has helped us to bring down uh the overall time to market and create models that are almost uh at a phase where we can say 7 60 to 70% of our data or the documentation needs are being autogenerated. We do not have to invest people. We do not have to invest that kind of a time that was previously needed. As I said 100% of the audit readiness in that 70% whatever we are doing all of that does not we just have a human in the loop to make sure things are not uh uh going hayware but most of it is autogenerated and that is 100% audit ready whatever the documentation automation that we have achieved using data IQ and data bricks that has helped us so you have seen the use cases, you have seen the platform, you have seen the architectures.
So what does it mean for us in the long run in at a very high level, right? So it helped us these platforms really helped us to reduce the insight uh time to insights by uh 50%, we are able to run some of these reports u and some of these models without worrying about the next set of hardware or without worrying about the concurrency or without worrying about the overlap of process. So we still have scenarios on in our data centers where we run uh sequential jobs because we only have limited capacity and all of that has been resolved and uh it has also helped us to bring the business teams and the tech teams and the our risk model risk management team closer because there is no friction now.
We we work very collaboratively. We are we are seeing that people tend to flourish when they have options and when they get uh multiple of these platforms which are uh which they do not h need specialized skill sets which do not need them to depend on someone else to fulfill their timelines. There was a scenario where people were used to uh wait for other tech team to create a dashboard or tech team to help them create a model.
Now the business team can start doing these things on their own. They have become self-reliant and independent. And from the tech side we have the balanced governance. Uh we have balanced the agility with governance because we know that we have processes in place. We have uh the uh data segregation in place that requires us to govern these platforms very well. So if we are able to achieve analytics scaling analytics without compromising on the compliance is the biggest win for us.
So as is the case with every single project that we have implemented right we always will have lessons that we can learn. What do what do we learn from all of this and how do we make sure that uh we carry forward those lessons is as one of the biggest things is we were able to implement or leverage no code capabilities of data IQ and we did not add anything on the risk side. If you look at the balance of it, we were able to expand our feature set, expand our capabilities, but at the same time keep every all kinds of risk to a minimum.
That's number one. And number two is uh the overall collaboration between the teams. Now people can start creating a workflow or a recipe that they can share uh with the global team. So we follow the sun model where the team comes in uh in Asia and are able to pick up from where this uh uh North American team has stopped and they can continue the work and they are able to follow the sun model. They're able to understand each other's work because most of it is uh visual UIdriven and uh they are now able to even uh share the clusters that we are creating on data bras or even have their own independent clusters depending on the criticality of the project.
So previously we we have known scenarios where people used to talk to each other and say hey I'm running this job please hold all all the other jobs. So that is a thing of the past. Now people are able to have their own independent clusters where they can run their jobs. It also helped us to kind of achieve a cultural shift. Okay. Uh as I said as we discussed it has empowered our users to do more and do it quickly. you don't you don't need to depend on anyone and it also helped us to bring uh from silos into a collaboration model.
So in the interest of time I'll uh just uh move on. So what can you do in your environment? How do how do you take this model and replicate it? Right? So this is something uh which is which is really close to my heart and all we believe at Morgan we learn from each other. we learn that's the whole purpose of coming to these conferences and all we learn from each other's experiences. So definitely this is something I would recommend.
This is how we started. We had uh we had conversations with our peers in the industry. We understood what are the techn what are the technologies that are available and how do they fit into our ecosystem. That's number one. But then start with identifying your own silos. What are silos? So we I discussed all the silos that we had but yours could be a different scenario. But definitely there would be silos. there would be a duplication of effort.
There would be reinvention of the wheel happening in your teams. So identify those and try and standardize it. Try and bring it into a standard on how it can be avoided. So I'll give you a quick example here. So during modeling right there is something called as feature set generation. I I'm sure most of you are aware of it and use it. So all our users used to do their own feature set creation. they were they didn't even talk to someone sitting next to them to see if someone has already created a similar feature set.
So that was the silos we were talking about and now what we have done is we have brought it all into one feature store where we said you don't nobody else gets to create feature stores other than one central team. You all can subscribe to features you can reuse the features and avoid your own pipelines. And that helped save close to 4,000 hours of manifford per year. So that's the kind of impact you can create by moving the small needle a small way to the right.
So that's something you can do and find a pilot that you can start using or test it out quickly. So most importantly test quickly, fail quickly, restart. So that's a that's the most important thing and we have done we we we had challenges we had obstacles. So that is where you need to create smaller goals smaller objectives achieve them move on create create multiple milestones rather than having one big target create multiple milestones.
So after you complete milestone three if you fail all you have to do is restart from milestone three. You don't have to go back from scratch. Okay. And last but not the least, as I said, join the community. So, we all are here. We all are learning from each other. Uh these are opportunities for us to share each other's success stories and learn something. Uh thank you very much for your time. I really really appreciate you all making this journey from uh St.
Masuni to here. I really appreciate that and I'm available here for any questions. Thank you.
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.