Sheikh Shuvo (00:00.854)
Matt, thank you so much for joining us. Really excited to dive into your world.

Matt McMullen (00:05.088)
Sheikh, thank you so much for having me.

Sheikh Shuvo (00:07.03)
Yeah, absolutely. So the very first question I like to kick things off with, Matt, is how would you describe your work to a five-year-old?

Matt McMullen (00:17.371)
I used to work at a company called CloudFactory, and it was a lot more fun explaining what that company did to a five-year-old or asking that five-year-old what they think we did at work. So, the preprocessing and data prep world, how do you explain to a five-year-old? Maybe it's like setting up stations in a kitchen or cooking dinner. You've got to clean your space, get it organized, get all the right tools.

Sheikh Shuvo (00:22.148)
Hahaha! Yeah.

Matt McMullen (00:46.019)
You'll first ask everyone in your household what they need, what they want, what they're desiring for that dinner. And then, ultimately, make sure that you have the right ingredients, get them washed, prepped, cut, get the water boiling on time, get everything staged and well-positioned so that everything comes out well. And basically, it's the same thing with prep for data before it goes into a computer. On the backside of that, something that will probably come up in this conversation, is data sum.

It's actually a pretty good analogy to use the kitchen workstation because part of what we do at work, data plus compliance, the compliance side is providing some sort of sticker about all of these ingredients that we've worked with to create this meal. Like how did we prep it? Where did the food come from? How much did it cost? Everything that went into the actual meal. So yeah, that's what we do at work.

Sheikh Shuvo (01:43.638)
I love it. It sounds like the, uh, the French concept of mise en place applied to the data world.

Matt McMullen (01:49.752)
Yeah, hopefully that's a good analogy.

Sheikh Shuvo (01:52.791)
Cool. Well, Matt, looking at your background, you started your career in the private equity world. What led to your transition into tech and working on data products now?

Matt McMullen (02:04.855)
Yeah, so I don't think I'm smart enough for anything like that to be on the finance side of it, the PE side. Um, I was definitely on the operations side. So I kind of always lived, even at a young age, in a world of people possibilities, and, um, early on in my career, I met this founder, this owner of a big holding company, well-known in the region that I grew up in, but ultimately took me under his wing. Taught me the operational side of PE. And I got to work with a number of companies underneath the umbrella, up to 20 something companies at one time. And ultimately, we would transition these companies, stage them ready for sale, but it was always a people and operational challenge. So transitioning into tech, I don't know if it was necessarily a finance to tech type of transition

as much as it was a transition out of working with 20 something companies at once and somehow having the bandwidth to live and breathe and believe in these companies and what they're all about, and then to go into working just for one and then having the dedication and the interest to again live and breathe that one company. So I think that was mostly the transition in that people possibilities and the operations change management really got me ready for a career in management and leadership.

Sheikh Shuvo (03:37.518)
Makes a lot of sense, especially with your background in psychology and research too. It seems like the best data set to work with at that point in real life. Oh yeah, along that journey then, Matt, what was your first professional experience with the AI world? How did you get started there?

Matt McMullen (03:47.122)
Yeah. Um, so I'm sure AI was lingering behind the scenes of our intranet, uh, back in the day or some sort of, you know, uh, enablement technology, but in terms of working for a company that specifically sold AI, um, or sold services into the AI industry, that didn't happen until a little bit later. Um, but right before that happened, I was working with a company called Big Belly. Um, probably the worst marketing company you've ever seen your entire life.

Sheikh Shuvo (04:26.222)
Yeah.

Matt McMullen (04:27.463)
But you've definitely interacted with our stations. It is the largest smart waste and recycling company in the world and stations on every single continent. 30,000 stations across New York City, for example, saving the city of Philadelphia $3 million in recycling and trash costs. But basically, that company built a platform for smart devices. So if you remember, smart devices were the thing, you know, cityscapes and making sure that devices could read everything from pollution to noise detection to pedestrian traffic to like smart parking. Well, we took all of those smart devices and put them in this lockable, robust trash can.

Sheikh Shuvo (05:15.554)
The IoT revolution of a couple of years ago.

Matt McMullen (05:24.223)
Made it solar-powered and the trash can compacted five times. So the reason why it was called Big Belly is because the founder of the company was obsessed with seahorses and there's a seahorse called a Big Belly and it can take on five times the amount of its size for food. Anyways, so again, horrible marketing. Fun fact is that we actually recruited and hired the Segway marketing team, so it got even worse. But anyways.

The company did really well and my job there was to figure out how to take that company and transition its GTM strategy into the private sector, which was truly difficult. It was always selling in municipalities, improvement districts, university campuses, all these companies that would eat it all up. But my first step into the AI sector wasn't until my now wife and I, we moved down to North Carolina. She was at Duke Law School and I wish I had a like a beautiful answer for this.

Sheikh Shuvo (05:53.942)
What's it?

Matt McMullen (06:21.463)
I've listened to some of your other podcasts and most answers have to do with like, well, I did this in college, did this at university, then did this, you know, and then this, and then, you know, I was always in AI or something. Anyways, but my step into AI just, it's pretty shallow. I just, I Google searched the top five startups in the Southeast, applied to all of them, and met the people at a company called Cloud Factory, fell in love with their product, their mission, but more importantly, the people there. And I remember realizing if I don't get this job, there's nothing wrong with them, there's something I need to improve. So I worked my tail off to make sure I could get that job and work there and have a semi-long career, especially for a startup there.

Sheikh Shuvo (07:07.938)
That doesn't sound shallow at all. It sounds strategic and technical. Well, as you started working more and more in the AI world, one really interesting metric background is also you've taken a really active role in the NGO world as well. I mean, two of the experiences or the projects that you worked on really stood out: your work on Next Step and Africa AI.

How did you get involved with those, and how did that differ from the work you were doing at CloudFactory?

Matt McMullen (07:43.951)
That's a good question. So the outsourcing world has a lot of traditional use cases and reasons for outsourcing work. The data vendor world, a subset of that outsourcing world, sometimes referred to as BPO. But the data vendor world hasn't really been around for too long. It started really gaining prominence and media attention, especially in 2015.

Companies like Sama Source, Cloud Factory, and iMerit popped up and really applied some sort of impact angle to outsourcing to take on the incumbents, to take on the established companies in the space. Next Step took a different twist. It used to be called Stepwise, and we transitioned it into a 501(c)(3). What we did at Next Step is we built a coursework platform for aspiring data analysts. The entire strategy and hypothesis was, okay, there's Snowflake, there's Databricks, there are these massive companies around the world who are offering incredible technologies. However, there's very limited resources overseas who actually know how to use these resources. So what happens if we match up that talent, or aspiring talent, with companies who might need access to that talent, to diversified talent, or in some cases, just lower cost talent? So we built a coursework platform that would act like a marketplace. One is where Databricks and Snowflake can build coursework onto this platform, and they could see how many people were taking their classes and graduating and then seek them out for employment. But then, you know, there's a natural demand cycle involved in that.

Sheikh Shuvo (09:29.408)
Oh, I love that.

Matt McMullen (09:37.547)
We ended up working with the MasterCard Foundation. They had something over like $1 billion of investment in Africa at the time on job creation alone. So they helped us fund the coursework. Next Step is still around, but as a 501(c)(3), it's obviously moving a little bit slower.

Sheikh Shuvo (09:57.09)
Hmm. Oh, awesome. Well, tell us about what you're up to now at Cogito Tech.

Matt McMullen (10:04.439)
Yeah. So, Cogito Tech, it's fine. Cogito is similar to many of the data vendor companies in the space. Many companies in the data vendor space have evolved to focus on a core competency or a core use case. But ultimately, there's still a tremendous amount of data work that needs to be done for AI companies. So if you think about it, like every day, especially over the past year,

Sheikh Shuvo (10:06.967)
Juneau.

Matt McMullen (10:34.787)
I'll give you another example. Before the past year, if I were to ask my parents to name an AI company, they might say IBM. Well, mom, dad, can you name something that might be smaller, like an actual product? They probably couldn't, but up to about a year ago, they would probably name OpenAI or ChatGPT. So every day these headlines are trumpeting some sort of new and astonishing development in AI.

Sheikh Shuvo (10:56.391)
Right.

Matt McMullen (11:03.179)
Whether or not it's Alexa, Siri, or GPT-4 or something like that. But one thing that's often overlooked is this unsung hero fueling the development, and that's just like the vast workforce that's annotating data, ensuring that the algorithms can even work. So, Cogito Tech has taken that focus and is making it useful.

So yes, we are helping these companies with data curation and labeling services, but right in return, we are also applying some sort of compliance angle to it too. So because there is such a vast workforce of unsung heroes, we're talking about billions of people. How do we show who these labelers are, who these workforces are from a demographic, from a pay perspective?

How do we pull back the curtain? How do we open up the hood to really show off their abilities? But more importantly, how do we establish embedded best practices in the dataset for these companies? And then the ultimate goal is how do we have some sort of nutrition facts for somebody who's interacting with AI? So Cogito Tech is at the forefront of that. There are a lot of big companies who are also addressing those challenges, such as the Data and Trust Alliance, which is 20 enterprises right now.

It basically mirrors exactly what DataSum is, and it was just announced in late November. And Cogito Tech is data plus compliance. So data labeling and curation, plus this idea of how do we show the compliance, how do we show the governance that a company like us needs to do to be able to create these datasets.

Sheikh Shuvo (12:45.506)
For those compliance and governance frameworks that you're applying to the data project there, where's that drawn from? Is there a central authority of best practices that you're basing that on, or is it something you've developed in-house?

Matt McMullen (13:05.355)
Um, so it's a good question. It's drawn from purely my frustration, and I'm sure many other companies' frustrations. Like I said, I've been in this impact sourcing data vendor world for quite a long time, but there were a lot more inspirational leaders in this space who have really been trying to drive a lot of change and make meaningful impact on the ground in these emerging markets for entire generations of workers, entire regions of workers. And I'm only just one person who's trying to make a meaningful difference as well. My frustration started and has grown over the past several years because impact doesn't sell. It doesn't matter what company you're interacting with. Just as an example, like when a salesperson's sliding through their slide deck demo and they come across, "This is how many people we've been able to employ in the past year and this is the impact that our clients are having." It's a feel-good moment for everyone involved for sure. And it's definitely a checkbox for these companies to have each year within their diversity and inclusion initiatives. Of course, it is. And those are great things to strive for and to have on your radar and your focus, but it doesn't sell. There's no way a company will be spending millions of dollars with the leading of their buying behavior being because they want to have an impact. It all comes down to cost. Whether or not it's timeline speed quality, it all comes down to cost. So my frustration has been because of that. So how do we not try to force a sale of impact? How do we just make it useful? Your question about what is driving this? It's both that, but also there are plenty of companies and organizations who are also focused on ethical integrity and data governance and in providing benchmarks and principles and that type of thing. One is GDPR. I'm sure that every organization within tech is familiar with it. There's also international laws for workforce well-being. But we also have an open contribution model for DataSum. So DataSum is five certificates. And the certificates, I won't bore you with all of them, but they include ethical integrity certification. So the bias levels within a dataset, the fair use assertion versus copyright, and the workforce well-being, like what are the employment terms of the team? So those are just two certificates as part of DataSum, and this is all open contribution. So all of our clients and the community at large can contribute to advancing the DataSum program. And you'll see a number of other organizations, like I already mentioned, the Data and Trust Alliance, they're doing the exact same thing. They have this 30-slide, beautiful presentation of essentially what DataSum is and ask for contributions and feedback. It's like, that's exactly what we want. So when you ask for, you know, like, "Hey, what's your North Star?" I would say that it's like this open contribution model. So allowing the community to be involved.

Sheikh Shuvo (16:23.886)
I'm counting on a more practical level, looking at how your customers have actually been using the DataSum product, how has knowing more about the origin of the data impacted their internal AI projects? Does it have an impact on what the requirements look like, what the QA looks like, and I'm wondering what the internal impact is on the project lifecycle.

Matt McMullen (16:59.663)
Great question. The impact is unbelievably inspiring. Most impact when it comes to compliance and policy and that type of thing, I won't even talk about regulation right now, but the policy and governance and programs and that type of thing, can be a headache. It can be a challenge for a startup knowing that they're going to move slowly always because they have to jump through hoops and get all these approvals. But every company that we've been interacting with, especially over the past year, has been super intrigued with the pre-processing policy and data governance. So as an example, a medical AI company will come to us saying, we have 1.something million words that we need named entity recognition work on. And we say, okay, great. Here's the pricing for it. And then they'll say, but hold on, we actually need five times that amount. Is that something that you can provide? Yes, because we also provide data curation. So we do two things. One is we have a dataset from the client.

And we're also tasked with going out and creating data from either other third-party private companies or public sources such as data.world or something like that. We need to both advise and consult the client on what fair looks like. So when we take a look at their dataset, especially if it has to do with population data, we ask, what's the outcome of the AI or the outcome of the product you're trying to create? And have you thought about those terms? Is your dataset representative of that desired outcome? So that means we need to go out and find data that comes close to matching, or where we have the ability to balance out that fairness. We're talking about different demographic representation in the data, whatever they need their AI to represent to have an ideal outcome for, is something that we can advise even before the data labeling stage. Then something that often happens, especially at the enterprise level, is that enterprises only pay for deliverables. They don't pay for effort or anything else, just the work that gets done on schedule, and then they issue purchase orders throughout the engagement. Fine.

With the enterprises that we're interacting with now, when it comes to DataSum, we apply DataSum to the deliverable so that they have a two-week snapshot on everything from a compliance perspective that went into that data, that went into the effort to create that dataset. All the people behind it, the entire workflow, how it was mapped out, the firewall protocols, and any issues with that. Everything is notated in that. So it lives as metadata for that dataset.

So it's changing a lot from the very beginning of the actual interaction with the client. And all of our clients are super intrigued and involved rather than them just thinking, "Oh man, this is just another hoop I have to crawl through."

Sheikh Shuvo (20:34.25)
Are there any particular customer stories that you can share?

Matt McMullen (20:40.503)
Uh, um, one of the largest companies in the world, basically what they needed to be done is, uh, there's OpenAI, there's Google just got in trouble, Meta is getting sued for copyright infringement within the training data. Well, this large enterprise said, "We understand the realities of copyright and fairness, fair use, and we need you to go and build a workflow according to these specs for this project and we need to issue a legal opinion on your fair use assertion." A lawyer, my wife's a lawyer, her listening to this explanation, she'll be thinking like you butchered that explanation, but the point is my job as head of corporate development is basically to go out and write that policy. So I had to write the fair use assertion that the way that we were collecting data from certain media platforms, for example, was within fair use. And here's the entire explanation on how. And it's basically an acknowledgment and guarantee that that's how we did it. So that when this big organization is then looking at the dataset, like wondering how this was put together, they can see that fair use assertion.

But even before that, then their legal team is looking at the fair use assertion and providing a legal opinion, meaning a signature saying that they agree to it. I think that's probably the key differentiator between a typical engagement over the past year compared to before this past year. And I also think that's a differentiator between what we're doing, what I'm trying to do here, and what any other company is trying to take on.

Sheikh Shuvo (22:32.434)
Oh, it's great to hear that the level of involvement is spreading beyond product and engineering and having leaders from all parts of the company at the table there. That's a huge value and a win right there. Well, to throw a hypothetical at you, Matt, so looking at the world of data annotation, some studies show that the average worker makes like about two, three bucks an hour. Now, if we were to apply the highest level of US labor law, and let's say restricted all US-based companies for any training data they had to use, US labor paying a minimum wage of $15. What do you think the impact of that would be to the AI industry as a whole?

Matt McMullen (23:29.475)
So basically, if there was some governing organization that said, it doesn't matter where in the world these folks live, but we need to raise the standard of the minimum wage for this type of work, something along those lines, and how that would impact. I can tell you right after that, the only companies that would be able to afford this type of service would be the major corporations. So immediately, any large to small startup.

Sheikh Shuvo (23:43.254)
Yeah, exactly. Yeah.

Matt McMullen (23:59.583)
Any large company to a small startup would have to look for an alternative. This might be a good thing though, because we're already seeing advancements in pre-labeled data or automation within the labeling process. There's nothing wrong with that. We're not going to put our industry out of service. What you're basically doing is demanding that teams like ours, the workforces that we're employing around the world, that they apply higher-level skills, that they adapt to the changes and use more creative thinking. Why wouldn't we want that stimulation on the team? So immediately, the big change would be only the big corporations could be able to afford it, and that means that the cost of AI development would go up, of course. There's another aspect too on the other end of what we do, the economic and social increase or could have a positive social impact like providing higher wages and better working conditions for the teams on the ground in these emerging markets. It would definitely be an uplifting component in these regions. But there's also something else. There's, how do you guarantee that the companies actually do pay that amount? Is there a reporting mechanism? But also is pay the only way of making impact, and you know, one sense having more money in people's pockets in these emerging regions, sure, but that's just one aspect of impact that and I think many of the other ones are often overlooked and never talked about in the media. So to answer your question, AI development, the cost of it definitely goes up. The economic and social impact, no doubt, like there definitely would be an uplifting in those regions. I think it might be temporary.

People do know how to spend their own money, but at the same time, there's other ways of making meaningful impact as well. So I would want more attention towards something else rather than just, you know, increasing that wage.

Sheikh Shuvo (26:27.503)
Now it's a super thoughtful answer. Looking more broadly as you think about the overarching strategy of your company and which way to evolve things, outside of tech, is there any industry that you draw inspiration from?

Matt McMullen (26:57.283)
Any, of course, you know, there are those, there's definitely canned answers to a question like that. So can I, can I say something that's not like an industry, but it's kind of from the, you know, conventional or traditional sense? All right, cool. I would have to say one of the most well.

Sheikh Shuvo (26:42.954)
You can interpret it however you'd like.

Matt McMullen (26:57.283)
The industry that I think has the most contribution outside of tech, and what I mean by contribution is from those within the industry, people who are preaching best practices, asking questions for clarification on those best practices, writing books, podcasts, you name it, it's unbelievably huge, is sales. So from a non-traditional industry perspective, I would say following sales thought leaders is actually really interesting.

It's everything from emotional and personalized functions of It's it's this emotional and personalized plus like this You know, um this functionality aspect of any initiative or any type of sale? And all of these thought leaders are talking about how to bring those things in like how do you How do you create a functional sale? How do you weave in personalization, how do you weave in this emotional aspect? And it's this cycle of like these really, really thought-provoking, articulate, of course, thought leadership, it's really neat. So I would say the sales sector is definitely something where I'm drawing inspiration from. That also has to go to do with this world that I try to live in with this, like people possibilities, you know, that's all sales is, it's too. people really trying to do something that's extremely raw and tangible, which is exchanging some sort of product and funds. So yeah, the sales sector, I also think that something that's close to that would be maybe like a user interface, like if that was a sector, so like a UI. Again, the same thing in terms of the sales sector.

Thought leaders speaking about their craft and what creates amazing UI and the things that they've learned from bad UI and building those teams and launching and shipping successful products and what that looks like and all the pain that went into it. That's fascinating. There's nothing more useful than taking those learnings and applying it to really any type of change management model or any type of a new initiative or any type of GTM strategy or any type of new team and project, like those takeaways are golden.

Sheikh Shuvo (29:51.032)
Are there any particular sales leaders you find yourself gravitating towards?

Matt McMullen (29:34.963)
Yeah, I wonder if they're gonna love that I called them out. Have you ever heard of sales assembly?

Sheikh Shuvo (29:41.526)
I've not, no. Tell me about it.

Matt McMullen (29:42.955)
So sales assembly is completely organic. They've grown so fast. I first was introduced to them in Chicago where they started when I used to live there about two and a half years ago. And two best friends, unbelievably creative, hilarious people, both with a strong sales individual contributor type of background. They wanted to create this uh, weekly morning breakfast for sales folks across the city. And when I first joined them, like breakfast, like, why would I ever want to do that? It's, it's really because the people who are joining this company have no other time. I mean, sales is the hardest high paying job, the easiest low paying jobs. So the folks who are going to be, you know, going to a happy hour and getting drunk and not being able to work their hardest the next day are probably the ones who, who don't have a good pipeline and quota. Um, but the folks at the, at these at the breakfast were incredible. Great leaders across the city and they've ballooned. They've grown so fast, it's all organic. Their thought leadership is both spot on and useful and there's a lot of tangibles that you can really, there's a lot of things you can take away. But more importantly, it's really entertaining. Like their banter, everything about that company, the way they hire, it's great. And now they're turning from a true consulting service for sales, like best practices and like this is how you like they would have like this hiring branch and like this hiring you know these hiring events that they would put on instead of just providing services they're actually moving more to platform so all of their advice over the past years is now more lives more evergreen and within these recordings and these modules and then they also have the personal touch too but it's incredible to see their growth.

Sheikh Shuvo (31:40.142)
Oh, that's a great recommendation. We'll definitely check it out. Looking at your own, own sales cycles then, um, a big recent news has been with all the, um, you AI acts, um, becoming finalized soon, have your customers started talking about that? I'm wondering what the impact of it, if any, has been on current sales cycles.

Matt McMullen (31:54.851)
Yeah. Everyone's talking about it. Um, everyone's talked about the, uh, the Biden, um, Harris, uh, AI executive order. Uh, it seems like the EU is moving extremely quickly while regulation. I won't really go there in terms of like what that does to innovation. Like there's, you know, I think it's hard to argue with the fact that. Regulation does slow down and impede innovation, but is it a good thing? Sure. Uh, the reason why I think it's a good thing is because.

Sheikh Shuvo (32:14.37)
Mm-hmm.

Matt McMullen (32:33.287)
I know this wasn't your question, but I want to make sure I say it. You know, these calls for regulation are a fantastic thing, especially for companies like mine. It basically takes this the biggest loud speaker in an entire region, entire continent, and it brings attention to something that's been underserved and has been hidden for so long, which is the pre-processing data prep, data labeling, data curation side of AI.

And it starts there, which is fantastic. Most monitoring and, yeah, most compliance platforms in AI don't come down or don't go up to the pre-processing side. They all stay within this ivory tower of AI deployment. They never come, they never go to the training data. So for these big government entities and groups to draw so much attention to it. It's a fantastic thing. But from a client perspective, in terms of our sales cycle, it's emphasizing that the sale needs to be more consultative and involved than it used to be. It used to, again, it used to always be about costs. Now it's more of how do you do this work? How do you advise us to do this work? That's the question all of our prospects, big and small are asking. So it's not necessarily slowing it down, but we have to become experts within the regulation, but ultimately it just comes down to data sum. So the EU act, once it's formalized next year, I think it's going to be early 2024, we'll take all that into consideration and see what we can change and adjust and better within data sum. But as far as like what our customers and our prospects are saying, they're saying that they just need more traceability and transparency within the process. We basically say, great, we thought about this five months ago. This is data sum. So it's just, ultimately, it's a lot of validation. It's incredible. We're really happy to be in this position.

Sheikh Shuvo (34:37.825)
Yeah. It's years and years of effort going into overnight success then.

Matt McMullen (34:46.379)
Yeah, it's incredible.

Sheikh Shuvo (34:48.13)
Yeah. Well, about the very last question I have for you is that, um, knowing what you know now, um, going back in time to your earliest exposures to the AI industry and making a career out of it. When you were interviewing for cloud factory, what type of advice and perspective would you give that version of Matt in the past?

Matt McMullen (35:16.408)
Uh, I actually think that the advice I would give is something that, um, I want to, I want to say it's the same thing. I try to encourage myself to remember every day, but I mean, I think that kind of sounds too superficial. So, um, just to answer your question head on, I think it's something along the lines of like, from a personal perspective and a personal development, um, I would remind myself, uh, keep your head down, do great work and your time will come. Like it's this, I am naturally an impatient person. I am attracted to urgency. Like I love when people find importance in every single thing that they do at work. I think it's incredible. I wanna be part of that team. But that often means that, how do you slow down? How do you have more perspective? How do you consider all the alternatives? Those types of things. So going back to the early stages of my career, I think that would be the number one thing.

I mean, just as an example, I would have given anything to be a fly on the wall in every single meeting. It doesn't matter what it's about. Because all I want to do is learn from like, how is that meeting run? How is that feedback given? How was it received? How did that argument go? Did that person get really mad? It's not just about the drama. Of course, drama and during meetings are cool. But it's really about the... I just want to know how these leaders are running these types of meetings and initiatives and decisions Ultimately, it just comes I think the advice is again just like keep your head down. Do great work and your time will come.

Sheikh Shuvo (38:30.387)
That's sage advice. If one of your sales cycles goes south, just keep that in mind.

Matt McMullen (38:35.503)
Yeah, absolutely.

Sheikh Shuvo (38:37.198)
Cool, Matt. Well, thank you so much for sharing about your background and your perspective. This has been super fun.

Matt McMullen (38:39.779)
Thank you, Sheikh.