Driving financial freedom with data

Tarush Aggarwal (00:00)

to another episode of People of Data, where we get to highlight the wonderful people of data. These are our data leaders from various industries, geographies, and disciplines, discover how companies are using data, their achievements, and the challenges they face. Today, we are welcomed by Maddie Daianu who has a stellar record in building high-performing teams and driving business growth through data and AI, currently leading a 100-member-plus team at Credit Karma.

‍

Maddie's career journey spans roles at companies like Facebook, Intuit, and The RealReal, where she's tackled challenges like building data teams from ground up, data literacy, and growth to millions of users. Welcome to the show, Maddie.

‍

Maddie (00:42)

Thank you for having me. It's great to be here.

‍

Tarush Aggarwal (00:44)

Are you ready to get into it?

‍

Maddie (00:46)

I'm ready, let's do it.

‍

Tarush Aggarwal (00:47)

Awesome. So what does Credit Karma do? I'm sure everyone knows that if they don't, they don't have credit scores in their country. But for those of you who don't know, what does Credit Karma do and how does data play a role in helping Credit Karma do what it wants to do?

‍

Maddie (01:04)

Absolutely. So as you said correctly, Credit Karma was the first company to provide consumers free access to their credit scores. Now, while most people know us for this free access to credit scores, we do so much more. Our founders always saw credit scores as a means to an end. When members check their score, they're typically doing it in the pursuit of something more impactful, whether they are taking out their first credit card or buying their first home.

‍

Credit touches really every aspect of our financial lives. So this is why you will find marketplaces, tools, and experiences across credit, loans, mortgage, and insurance, as well as checking and savings all within the Credit Karma app. Having said all that, our vision is much bigger than that. We want to be a comprehensive, self-driving financial platform.

‍

that propels our members forward, no matter where they are in their financial journey. to be able to do this, as you would imagine, we have to really harness the power of data. We have financial data from 140 million members, which when combined with machine learning and AI, has been very powerful in providing our members with the product offerings and tools to manage their financial lives. I joined Credit Karma about...

‍

two years ago at a very pivotal time. And that was the advent of Gen AI that today powers various experiences on our app. So Credit Karma is a place where people can come and get their holistic view of their finances. now with Gen AI and the reasoning capabilities that we get from it and the contextualization, we are truly able to connect the dots between what our members want to see when they come and land on the app.

‍

as we guide them through navigating their finances. So as an example of this, just briefly, we launched not too long ago something called CY, which is a feature based in Gen.AI that's attached to your credit card provides an AI-generated explanation as to why we are recommending a certain card to a member. So this includes the current state of their wallet, card benefits, and helps them match their spending habits.

‍

Features like these give our members the personalized information they need. It's grounded in their own financial data and it's helping them make informed decisions.

‍

Tarush Aggarwal (03:30)

Yeah.

‍

That's awesome. It's very interesting that you're one of the brands where data, in some ways, is your product. So being on the data team plays a doubly interesting role. What's the culture like across the company? And how does that map into the culture of the data teams, which typically tend to be a little bit different

‍

Maddie (03:56)

so data is really center of how we do a lot of the things at Credit Karma. And as you said very well, it is part of our culture. So let me walk you through maybe how data is embedded in our fabric at Credit Karma. maybe I can talk through that entails for us, but also some of the systems that we built that leverage essentially data. data has been part of our fabric from the beginning of time.

‍

But as you would imagine, data needs to be considered and be intertwined with AI and machine learning to really unlock meaningful value. data only becomes powerful when we are able to do that. And conversely, it reaches its full potential when we have high quality data. So as part of that, in terms of our culture, we have, from the beginning, even prior to me joining, a lot in being able to build

‍

recommendation systems that really propel that data forward, personalize our product. And we are making a lot of decisions with using a lot of data-driven capabilities, A-B testing, launching products that are well-grounded in understanding how our data influences our products and our new features that we are launching. So let me give you a bit of context about our data.

‍

We have right now more than 60,000 features about our members. As you would imagine, no two people have the same financial profile or situation. So we essentially have to be very considerate about how we use this information because no two people having the same profile means very different ramification for each individual. Because personal finances is so intensely personal.

‍

seeking a one size fits all advice is not enough and providing that is not sufficient to really meet our members where they are in their financial journey. So for instance, we are using curated data that we can help members with tell them where they stand financially. we are able to help them the right decisions wherever they are in that financial journey. instance, whether it is checking their credit card,

‍

score or their net worth or home equity and transactions. We are providing essentially that one stop shop so they can come in and check their finances. I'll pause a little bit there. There's a lot that essentially we educate our members with through the Fusion of With Data. Having said that, I do want to touch at some point either now or in the next sections on the systems that the team has built because without that, we will not be nearly as powerful.

‍

Tarush Aggarwal (06:32)

Yeah, no, we'd love to get into that. What's the data footprint of the company look like?

‍

Maddie (06:38)

So me maybe tell you a bit more about the systems and then I can walk through the data footprint and the teams. as mentioned, we have this vast amount of data about our members, than 60,000 features that depict the member wherever they are in their financial journey. But without essentially the systems that we build, we will not be able to really harness that power. So for instance, you might have heard of this, we...

‍

had something launched in 2018 called the Livebox platform that is helping revolutionize the shopping experience for members by giving them insights into offers that they have the highest certainty for being approved for. So Livebox is our proprietary enterprise platform that allows lenders to leverage thousands of de-identified data points from Credit Karma and to really help members provide them with better certainty as they apply for a credit card or a loan. But the piece that

‍

Tarush Aggarwal (07:32)

Yeah.

‍

Maddie (07:33)

I'm really pumped about is something that my team has built even prior to my time and we've been continuously optimizing is the recommendation engine. is essentially built on top of our machine learning infrastructure and it's intelligently ranking in a very tailored way to our members. So this ensures that we get in front of our members with the right offers, insights and tools helping them along with that achieve their goals based on their financial picture and behavior.

‍

so that's a bit about essentially some of the biggest, would say, investments that we've done from a recommendations perspective, from a system and AI buildup perspective that has enabled us to open up and have the time and the ability to invest in zero to one areas like Gen AI. So right now we are using Gen AI to provide that contextual insight

‍

to our members and to help recommend a certain offer, really without having building systems upon which we can really have a well-oiled machine, we will not be able to build Geniac capabilities or explore as many zero-to-one capabilities as we are doing today.

‍

Tarush Aggarwal (08:38)

Yeah, that makes a ton of sense. How about what stack do you use in order, know, sort of obviously in the data and analytics side, but also as you think about AI, Gen. AI, what's the technologies which are powering all of these capabilities?

‍

Maddie (08:54)

Yeah, so in terms of our stack, we are using Google Cloud. we have been using GCP services data warehousing and machine learning infrastructure for a while now. For both training and other components, we also have quite a bit of in-house build systems that leverage GCP capabilities that enable us to really evolve systems.

‍

I would say something that there's a unique piece that I found joining the team that is particular to Credit Karma, which is we build our systems in a way that allows us to really expedite getting models into production. So our training infrastructure has very little drift between the training environment and what deployed into production. So this means that we don't have to spend additional time productionizing the model.

‍

which really helps expedite how we put something in production, which has been very beneficial, especially for us in the last 10 years.

‍

Tarush Aggarwal (09:50)

And when you think about Gen AI, are you doing this yourself through vector databases and RAG? you using an off the steps sort of solution, either grounded or ungrounded? What's your take on this?

‍

Maddie (10:06)

Yes, absolutely. So in terms of Gen AI, have various systems that we build internally. So I will start essentially with something that Intuit as a whole is very proud of, is something called Gen.os. Gen.os is a full-blown operating system that supports the responsible development of Gen AI powered features across the whole technology platform.

‍

In addition though to GenOS, we, Credit Karma, as well as any other business unit that is within Intuit, has to build certain scaffolding can essentially enable these GenAI systems to fit our needs and financial use case. this includes essentially things like building machine learning models that do card rails and safety evaluation frameworks and ability to route responses, the ability to augment responses through Rags.

‍

amongst other enhancements. So we essentially have been able to build the scaffolding around GenOS to really be able to enhance and our use cases. I will say this though, that one aspect that has been quite interesting and might be of interest to this audience is that whenever you build a system scaffolding from an engineering perspective around commercial or open source LLMs that evolve quite a bit,

‍

that generative behavior changes from an LN perspective, then the system can also change and the response of the system can be fully be swayed. So being able to like strike that balance where have enough stability with respect to not only our systems at the space that the GenAI landscape is advancing while having that stability and responsiveness internally.

‍

has been one of the most difficult pieces as we have been investing in JMAI.

‍

Tarush Aggarwal (11:52)

That makes a lot of sense. It's exciting to see that this is something which is sort of coming from the mothership and something which all of the brands are able leverage. What is the size of the data team look like at Credit Karma? And if you have a central team at the parent entity level.

‍

Maddie (12:13)

Yeah, so as you mentioned in the beginning, my team specifically is a little bit over 100 people. And this is essentially across data science, machine learning engineering, various data platform teams, business intelligence, experimentation platform, all within the same umbrella. But we have almost 900 people team in engineering at Credit Karma that support essentially

‍

not only my team, but all the business units within Credit Karma accordingly. And then within, as you call it, the mothership at Intuit, we have our or chief data officer Ashok, who has over a thousand organization that is also essentially very central to how we work with them. GenOS, for instance, is part of his organization. This platform that we are relying upon as we build GenAI capabilities. And we work very closely across Intuit.

‍

his organization to support and advance our capabilities.

‍

Tarush Aggarwal (13:11)

That's very interesting. We typically see this sort of centralized architecture at single brand level, but to see it across into it, and then also they are building technology across all of the other brands. What do do when have a new acquisition or a new brand which might not be on the GCP stack or might not be able to go leverage some of the central technology?

‍

Maddie (13:37)

Yeah, I mean, right now, the that we are on is, as you said correctly, GCP. Conversely, Intuit is using AWS. So we have essentially already a multi-cloud integration that we are looking into. And this is not only prevalent as we are accessing GenOS, which is this platform upon which we are really building genetic abilities, but it's also prevalent as we think about something that I'm really pumped about, which is

‍

bringing our data together and building a unified consumer profile that ultimately what I believe will do is unlock into its biggest competitive advantage. And that is combining financial data from Credit Karma about our members, along with tax data from TurboTax and really having that view holistically that can depict not only where the member is financially, but also where they are from their tax perspective over the course of the years.

‍

will be one of the most powerful things that we can do as one company. And that has to be built within a multi-cloud construct.

‍

Tarush Aggarwal (14:39)

Yeah.

‍

Speaking of exciting projects, what would you say is one achievement which has significantly impacted the business which you are proud of?

‍

Maddie (14:51)

Yeah, so I've already touched on on Gen AI and we have been talking about aspects of the multi cloud system. I will say that very important for us to recognize that success for Credit Karma Intuit as a whole doesn't just lie in the quantity of the financial data that we have from 140 million members, but also this historical aspect of it. What I mean this longitudinal

‍

multi-year tracking of a financial journey for a member where we can start depicting what I like to call financial life events. And that is actually not easy to do. And we really have to like think about how we build our systems very intelligently to be able to extract that level of highly consistent, fresh, very relevant, timely financial life event aspect for our members.

‍

And would say, it's nearly impossible for any other company to do because we have 10 plus years of financial data, which we have been harnessing part of the recommendation system that I mentioned to you as well. But now is really the time to come together and unlock the power of this data combined with Credit Karma TurboTax that we can really leverage this holistic financial view combined with tax.

‍

and be able to really unlock this depiction where people are across a multi-year journey and be able to serve products along that journey more effectively.

‍

Tarush Aggarwal (16:20)

That makes a lot of sense. one challenge which you've faced in your journey which you're currently focused on solving?

‍

Maddie (16:29)

Yeah, there's many challenges and they relate both to one strengthening our data foundation, but also the Gen AI application. So as I mentioned to you earlier, we are using a lot of these capabilities from Gen.OS that enable us to really build capabilities for Gen AI internally. Having said that, one of the biggest aspects of this is really being able to not only keep up with a fast moving landscape.

‍

Gen.EI is very hot. It's something that everyone is investing in. Being able to identify product market fit for especially the financial landscape and domain takes a lot of innovation. But sometimes that innovation can be aback, especially when we have to spend a lot of time doing what I mentioned earlier, which is building this engineering scaffolding around Gen.OS system and around really the elements that are either available commercially or available

‍

source, whenever those shift and evolve, we have to like pause also shift and evolve our internal engineering support. So striking the balance between like how many engineering resources we throw at the problem while really being able at the same time to advance innovation and really nail that product market fit

‍

challenge that I think not only Credit Karma is dealing with, but likely a lot of the companies are dealing with.

‍

Tarush Aggarwal (17:56)

Yeah, I think that's something which is on everyone's mind. If we have any listeners in the audience who are interested in Credit Karma, interested in Intuit, what advice would you give them if they're looking at careers?

‍

Maddie (18:12)

Well, there's certainly a few roles that I would say I'm really excited about within my organization. In particular, I'm looking to hire a data architect that can really help us with I mentioned to you earlier, this unified consumer profile that I believe will be unlocking one of the biggest competitive advantages for the company and into it as a whole. also are looking and will be hiring GenAI specialists, so people that want to come and join.

‍

build the rocket ship with us would be fantastic. And we would love to learn have them apply for the roles. And there's a very big investment from Intuit well as our CEO, Sasan, and bringing in technologists Intuit. So I encourage everybody to take a look at the Intuit website. There's likely a lot of exciting roles beyond the ones that I just mentioned.

‍

Tarush Aggarwal (19:05)

Matty. Thank you so much for sharing your wisdom with us today.

‍

Maddie (19:09)

Thank you for having me.