AI and Machine Learning to Transform Your Workforce | Kian Katanforoosh

Welcome everybody to the Scale up show. This is your host, Ryan Staley. And I have a very special guest with me today. I have Kian Khataf Roosh. Kian is the founder and CEO of Workera, which is a groundbreaking platform that is reshaping the landscape of professional skills assessment and development across all parts of the business. Okay. Something really cool about Kian is he is a visionary entrepreneur, a Stanford professor, also a Coursera lecturer for machine learning where he's had over 3 million learners have taken his course that he's taught. On top of it too, he's a founding member of Andrew N's Inked Deep Learning. AI Keon, welcome. Happy to have you on the show, man.

Thanks, Ryan. Excited to be here?

Yeah, dude, I'm excited, man. We're going to nerd out on machine learning today. We're going to nerd out on skills assessment and everything that you do. But at the same time, I really want to get into your background because you have a, a unique path and how you got there. Before we do that, I want to do what's called the revenue rundown. So where are you guys at in the stage of your journey for your ARR?

We are roughly in a single digit million in ARR approaching hopefully double digits soon. And over the course of the last three and a half years, we've raised our series B most recently, which was our last round of funding at 23.5 million. And we serve about two dozen customers in the enterprise. Most of them are Fortune 500 companies and multiple governments.

Oh, nice.

As well.

Okay. So much, much focus on the larger organizations. If that's your pricing model. What's I guess like what's your primary go to market strategy for revenue growth?

It's mostly a land and expand model where we start with initially one group of employees, maybe hundreds of employees, typically within an organization, trying to demonstrate the value that we can provide to that group and to the executive or manager of that group and then expand within the same line of business or across multiple line of business.

Okay, excellent. And is that done through outbound or what's your method for customer acquisition?

It's been mostly outbound sales for the majority.

Excellent. And then what's your primary? I guess what's your team size right now? That's one thing I didn't ask you. What's your team size?

We have 80 people and we're distributed across 25 countries. Wow. Very global.

And then walk us through your solution in two or three sentences and who it serves. Just so everyone has an understanding of what you're doing and how you're executing.

Yeah. So we provide a skills development platform that can help companies understand their talents by measuring skills and then use that skills data to power talent applications. The first talent application is typically upskilling. So now that you understand your employees, you can personalize upskill skilling for each of them. But other talent applications include internal mobility, project resourcing, talent planning, workforce planning, and many more, such as recruiting or, or, or, or, or other, Other talent applications. Yeah, that's, that's pretty much the, the rundown in a sense.

Okay, so for like skill side, but anything from talent management to talent acquisition, is that what you say? Is that kind of a good.

So typically it would start with an executive having a large workforce under them that they're trying to benchmark, measure upskill, provide personalized learning to. And we help them get all of that skills data in an aggregate manner, provide feedback to every employee on their strengths and gaps, and essentially scale mentorship with an AI mentor across the organization on an ongoing basis. Typically, those companies are transforming. They may be at the beginning of an AI transformation, they may be actually in the middle of an AI transformation. We allow them to measure themselves year over year with an objective assessment that is computer adaptive and that tracks their progress over time from a workforce perspective.

I love that. Okay, I know you mentioned you're funded. Were you ever bootstrapped or were you funded from day one?

We were funded from the ground up. We raised the seed round in 2020, our Series A in 2021, and then our series B most recently.

Okay, and why did you decide to go that route versus the bootstrap route? Like, I always like to ask founders this because it's, it's interesting. I've gotten all sorts of answers, so would love to hear yours.

Yeah, mine was that what we want to achieve requires investments, especially if we want to do it as fast as we plan to do it. It's also, you know, we're building a company that is deeply technological, so there's a lot of upfront investments in product development and research and development. And that, that taken into account, it made more sense for us to take the venture back route than the bootstrap route it could have been otherwise. Yeah.

Okay, that totally makes sense. So let me ask you this. So on your way to building your company, how did you. I guess, what was the light bulb that went off to say, like, hey, this is something I need to do, and then go all in with a venture background? Because that's Usually the most aggressive in terms of growth and expectations that you could have. So, like, what was that? That moment in time and kind of what got you to where you're at today?

Yeah, so prior to work here, I was teaching AI classes. I created the Stanford Deep Learning class with Andrew Ng. I also helped him start Deep Learning AI, which teaches AI at scale. So we're talking millions of learners and everything we were doing, we were trying to do it as accessible as possible at scale, which also requires us to think differently. When we were teaching AI, a lot of our students, at least when I was connecting with them, were telling me that content was not anymore the bottleneck to develop a better career. They had access to classes across any type of AI skill or machine learning skill or data skills that you can think of. But the gap was in mentorship. And mentorship is a tough one to scale. Like, people feel lucky to have a mentor. When you find a mentor, you're so happy, but you shouldn't be. You know, you should be, but everybody should have access to a mentor, ideally that can guide them through the next stage of their career, that can understand their strengths and gaps. And so this idea emerged on, you know, we want to create a scalable mentorship. And then when you think about mentorship, you want to, you want to think about it scalably. So we break it down into pieces. The first step of a mentor is to understand someone's skills, otherwise you cannot mentor the person. So we decided to start building an assessment that requires knowledge of psychometrics, but revamped in the age of AI. And this is also a product investments upfront that needed funding. And then you also need to develop an ontology to help people set their goals. It is very hard for people to set career goals on their own. And so how can we develop an ontology that covers all sorts of goals at a granular level that can help people shoot for the stars. And then once you have an assessment and you can shoot for the stars, then you, you can put together a massive recommender system, essentially. And fortunately we, we've built as a community in AI a lot of good recommender system. You can think of Netflix, which is the popular one, and those recommender system with the right amount of data and the right data on skills can be used and leveraged for personalized learning and mentorship. And so these are, you know, the big pieces that needed to come together for mentorship to scale. And it was the beginning of workera.

Okay, I think that's a fascinating approach that you have for that, like, so let's break that down. Like, obviously the skill assessment is there because there's, there's anything from like Colby to disc profile. I know disc is more communication, how people communicate, but there's, there's Colby in terms of like who people are and their mo. There's strengths finders. Is it just like an integration of multiple layers of that to understand the skills or how do you kind of deconstruct that with a person? Because that seems like that would be hard yet there's some proven paths that have been done before there.

Yeah, yeah, that's a great question. You know, historically, assessments have been stressful. Like people have taken assessments for getting into a job or getting into a university or getting into a program, and they've also not been built for feedback. They were meant to select. So we had to change the mindset of assessments in reinventing essentially a user experience that is empowering, that is encouraging, that is mentor, like. So this is a brand new type of assessment that probably the world is not used to, just makes sense. We call that formative assessments as opposed to summative assessments which are meant to select. The second piece is assessments have been typically built pass fail. And so you either get in or you don't get in the program. Our assessments are not pass fail assessments. You know, you, you, you have to get feedback that is granular. You may have strength, you may have gaps, you may have areas that you need to focus on for your learning, areas that you don't need to focus on. And so we also needed to introduce new technologies to pass fail assessment to assessments that provide granular feedback. So for if I give you an example, our computer adaptive assessments uses technology that can predict skills that you may not test directly on. If I, if I ask you 2 times 2 equal 4 and you answer correctly, I can also predict that you can do two plus two equal four without asking you. I can predict that you can do 2 -2 +4 because I understand the semantic relationship between those skills and the connection between those skills. And so you extend that concept and you can generate predictions across hundreds, maybe thousands of skills in the matter of tens of minutes or an hour or two, which is unprecedented. And this is really a new category of assessment we're talking about, which is different from the disc profile or psychometrics assessments that you may have used or people have used in the past.

Okay, I think that makes a lot of sense. So let's go to the next one. So you're talking about like the creation of goals for, for like the career, which. Yeah, I agree. Like, I've interviewed numerous folks, hundreds of folks, and most of the time people don't know exactly what they want from their career. So what's your kind of approach to that?

So goals are interesting. I'm going to talk specifically about goals for workers in companies because, you know, my focus right now and the focus of workera is on the workforce, but that also applies to students in univers. And even if you go down to K12, it could, some of it could apply.

There is this new trend of skills ontologies, meaning companies are developing sort of skills maps that are meant to be used to help people define goals for their career that are skills based. Ideally, those skills ontologies are tied to projects. So you have projects, you have business initiatives, and then you convert those into skills, and then people will develop those skills, demonstrate those skills, et cetera, with assessments, with other. In other means, with other measurement instruments. Those ontologies need to be flexible and structured at the same time. So let me explain why they need to be flexible and structured. They need to be structured because most companies share a lot of skills in common. If you're an account executive at company A, an account executive at company B, probably 80% of the skills are common. You may be, you may have to do discovery questions, you may have to prospect, you may have to, you know, just do certain things. Product demos that are common to the two jobs. However, and that's why you need a structured ontology that is shared between those two companies, so that those two companies can say, we want to test on a standardized basis how people do discovery.

But there's 20, 30% that we find are different. Maybe one is responsible for the type of selling that happens in an organization versus another. This one is enterprise first, this one is SMB first. The industry that you operate in, the language that you use or the methodology for sales that you use. And that requires a flexible ontology where each organization can extend a public ontology with skills that are specific to their own use cases. And that's what we provide to companies. A mix of a structured ontology that has collected tons of data and benchmarks across companies and a flexible ontology that can be developed by the company itself.

Okay, so I'm following you there. That makes a lot of sense. You're basically applying Pareto's principle for the structured content piece and then the specialized version at the company level or uniqueness sounds like.

Right? That's right.

So, okay, so then let's go on to the master recommender system that you have.

Right.

Like, how do you approach it from that perspective in terms of like recommending, you know, having a system that recommends how people upskill.

So the, the recommender system will be as good as the skills data that fuels it. In other words, you can throw transformer based recommender system, which are best in class deep learning algorithms nowadays at poor data that is noisy and you will get worse results typically than a very simple recommender system on really rich and meaningful skills data. So the skills data is critical in order for the recommender system to work. In other words, if I know that you are 7 out of 10 in sales, that's one type of skill signal, it's very high level, it's not very descriptive. I will not be able to personalize learning for you with a good recommender system. Instead, if I knew that sales is 500 skills, each skill is tied to a specific cognition in the brain, a specific task that you do on a daily basis. And I have predicted those 500 skills through very rigorous measurements for you, then a recommender system can take that into account and say, hey Ryan, you have one research publication that will help you gain that skill. You have one YouTube video that will help you gain that skill. There's a podcast that covered those three or four skills right there. And now it gets interesting. And so it's all about the granularity of the skills data and the quality of the skills data that powers those recommender system. That makes sense.

Oh yeah, that totally makes sense. I mean, I've seen that just from experimenting with prompts and fine tuning and with OpenAI, it's like.

I get it, like the data is totally dependent on the type of prompt that you put in, which is like the data input that you're talking about for kind of the other side of it. So similar but different, right? Same kind of concept.

Yeah.

So let's progress a little bit. I think that that makes a lot of sense in terms of how the program or the solution, I should say, is designed and how you're kind of working with it. So I guess like, you know, can you provide some real life success stories and examples of how this has made a big impact on people's lives or organizations lives. So I'd love to hear that. And that's always where there's kind of the devils in the details on some of this stuff.

So yeah. So from a learner perspective first and then I talk about the manager's perspective. From a learner perspective, you are saving significant amounts of time. So in the past, a manager may ask a learner to take a 200 hour class and the learner will just do it because their manager asked, but will complain that it was not relevant to their job, it was too easy or too hard, and so it was an ineffective use of time. In 200 hours you can do so many more things if you focus on the right skills. So skills building becomes highly focused and intentional. On top of that, benchmarks allow people to understand where they stand. Sometimes you may be better than what you think, sometimes you may be worse. And you want to know because you want to be able to uncover your unknown unknowns and focus on them. So we have all the time learners telling us that I now know what to focus on and I did not know before. It also creates opportunities for people. You may have an organization with thousands of people and people are struggling to demonstrate their skills, show their skills to managers and get opportunities that match their skill set. Now it's different because everybody has a skill signal on workera. In certain organizations you can open up opportunities for them, you can retain them, you can make them excited about the latest technologies and producing results for the company. On the manager side, more and more pressure is put on managers to be talent experts. Like essentially your job is to run talent, to manage your talent base. And managers have never been trained at being talent managers. They oftentimes have no idea what are the strengths and weaknesses of their team, especially as you think, larger and larger teams. And so this tool can allow them to keep the pulse granularly on the skills of people at an aggregate level, anonymized or transparent depending on the culture of the company. There's all sorts of measures that can be put together on the platform. And it just makes managers able to focus on what they're good at and have sort of an AI assistant that help them be better at managing talent. Sense.

Yeah, that makes a lot of sense, a ton of sense, because it's so true. I mean like you have to actually do the job, but then you have to provide leadership to inspire people in your organization and then you have to upskill them, right? So they could either create a succession plan or fill gaps that are existing within your own personal org. So I mean that makes a lot of sense.

So. And oftentimes, yeah, oftentimes as a manager you don't have time to keep the pulse on what's the best content out there. It should not be your job to be constantly on content platforms trying to figure out what's the best article this week, what's the best video to study. And so you need an assistant to do that for you, essentially.

So what do you do? You take like a feed then of like the most relevant knowledge based on keywords or like, how does that work? Because that's really interesting.

It's based on skills. So we would. Our job is also to constantly measure skills and tag the best, index the best content out there to those skills so that if we identify that you're lacking a certain skill, we have verified content that is popular to learn that skill.

Does this work with like CEOs as well?

Yeah, CEOs can use it. Typically they may use it for their team. So they may define the ontology, what they want, and then we would do that for the team. But you also can put together a test for CEOs, you know, CEO of a series a company, like, what are the top 100 skills under that?

So do you use this for all your employees or just some of your employees? Or how do you kind of deploy it?

Yeah, you can, you can deploy it by groups. So you would create skills pathways for different sets of employees. So you would create maybe one for your account executives, you will create one for your customer success managers, you will create one for your engineers at different levels. And those pathways will include the skills that you care about for their projects.

Okay, so let's go more over to the business side now. I know we went super deep on your product, which was just fascinating for me, but like, let's talk about your business. So as you're looking to continue to grow, what do you have you seen over the last few years that's been the single best growth strategy that you've leveraged to grow the business to almost eight figures at this point.

Single best strategy.

I think identifying the right.

Ideal customer profile was important for us and then making sure.

That we provide value and thought leadership to those folks. So in our case, we work very well with companies that are trying to undergo an AI transformation. They also have a certain level of maturity on the learning or HR side. So we have this two by two map, let's say where AI maturity of the company and the LND maturity of the company. And if there's a sweet spot where people love our product, they just think that it's the next.

Tool to use in order to track AI transformation. Fortunately, over the last two, three years, a lot of companies are thinking about AI transformation, you know, and it's getting more and more urgent in people's minds. And so we are trying to, you know, provide good content Good advice. Good thoughts on LinkedIn or elsewhere on how to do AI transformation. Well, from a workforce perspective.

Well, I mean that's a. It's an awesome topic, I think, because it's very, very timely. So what do you think are the most critical components for an AI transformation? You know, let's talk about organizationally and then at an individual level in terms of how someone would approach this today to keep up.

Yeah, you. I'd split it, I'll simplify, but I'll split it into a good AI transformation requires strategy, technology and people. That's the three pillars. Strategy, fortunately can be put together pretty rapidly, assuming you have an executive that is very knowledgeable about the topic. Technology nowadays can be bought. Like you can buy data infrastructure, you can buy cloud computing, you can buy hardware, you can put together an MLOps, AIOps, whatever stack you want almost overnight. I'm simplifying, but within a few months the limiting factor to transformation has been people and the ability to keep people up to date in terms of skills and perform change management. If you think about cloud transformation as a proxy for AI transformation, just because we've been through cloud transformation over the last 15 years, AWS launched in 2006, GCP launched in 2010. So it's been like roughly 15 years. And still you see the majority of Fortune 500 Global 2000 companies in the middle of a cloud transformation. Still cloud engineers are highly sought after. Still people are trained on cloud literacy. And so our goal is to accelerate transformation through the people components so that AI transformations don't take 15 years.

That's funny.

Let's talk about that, Matt, because things are happening probably faster than they've ever happened before from an innovation perspective. So let's talk about this from a non technical person's perspective, right. Employee perspective or leaders perspective. How do you continue to upskill someone with AI when they're non technical with how fast innovation is happening right now?

Well, I think that obviously you don't train a non technical person the same way you train a technical person. So you will create a skills pathway that is different.

However.

I actually think of four groups in a company. Typically you have the centers of excellence. Those are machine learning engineers, deep learning engineers, data scientists, data engineers, all of that in that bucket. They are builders of the AI system, so they need to know the builder skill. There's a second group that is what we call AI X at workera. Those are not builders, they're consumers of AI systems. So they may be experts at something else and they're trying to Add an AI toolkit to their skill set in order to perform well. Just to give you a data point.

The class I taught at Stanford for six years has 2/3 of students not majoring in computer science, even if it's a computer science class in department of Computer science. But.

It'S not common to see that there are mechanical engineers, electrical engineers, aerospace engineers, medical school, et cetera. And they do very well in projects because they have the subject matter experts. So a medical student may try to build a chest X ray detector to detect pneumonia, for example, or a pathology. And they know what metrics they're going to use to evaluate the model. They know what the deployment environment looks like, they know what regulations they'll have to comply with. A computer science typically would not know all of that. So they cannot really push the project forward as fast. And then the last two groups to come to your question about non technical individuals, you would have people who need the level of AI fluency, I think because they interact with technical individuals frequently. So those may be recruiters that help hiring managers recruit. They may be sellers that need to understand the technical lingo, they may be marketers that need to put thought leadership pieces together with subject matter experts. And there you need communicating about AI. You need auditing AI system understanding what is AI, what is not AI. There's a set of skills that belong to the fluency level and finally the literacy level would be a lot of blue collar workers that interact with technology, interact with machines, and need to be able to understand how to leverage that technology for their career and their work. And those are also different skills.

That's a good breakdown. I really like that, the way that you kind of categorize them, because it is true. So do you think of all those four buckets that you mentioned, you seem like you're very strong on the domain expertise. Folks that have that rich domain expertise.

How do you suggest they make the jump from non technical to technical even though they have all that domain expertise?

Not that they need to build it themselves, but I mean, bridge the gap I guess you could say is probably the better question, a better way to ask it.

Yeah, My advice is actually don't switch careers. Is if you're a great mechanical engineer, you love what you're doing, stay a mechanical engineer and don't worry. Just study AI and bring it to your subject matter expertise.

I feel that AI is one of those technologies that actually can be picked up pretty easily by engineers, if you think about it. And the reason I say that is because we measure AI +X individuals in companies. So we work with one of the larger energy giants in the world. And they have a lot of mechanical engineers, they have turbine experts, they have electrical engineers. And when you measure mechanical engineers with a test, you realize that they're actually better than many AI people at linear algebra, at matrix multiplications, at dot products. Like, there are certain skills they just have because it's part of their mechanical engineering curriculum. They don't need to relearn that. So instead you can personalize learning for this group and say, hey, where your gap is is coding. There's a set of coding skills that you need to develop in order to be able to leverage those technologies. And that's why our technology is very well also positioned for AI X folks who have diverse backgrounds, different skills that can be leveraged in order to accelerate adoption of AI in their subject matter.

Love that. I'm going to check it out myself. You got me intrigued now. Totally intrigued. Well, it sounds like things are going pretty well with how fast you're growing. What would you say is the biggest challenge you've had with growing your business right now?

Yeah, talent. Honestly, like executive hiring has always been hard. We are hiring for executive position in the go to market team and, you know, finding the right person that loves education, loves talent, also is knowledgeable about technology like AI. It's always a challenge. I think all my founder friends have the same issue and would probably mention that other than that.

Talent is a. Education is a very.

Is a market that is very complex. There are so many tools in edtech. There's lmss, there's lxps, there's hriss, there's workforce planning tools. And so teaching the market what it means to be a skills tech platform. A platform meant to measure skills and provide skills. Data is complicated, so you need to, to really see the market mature before people can understand what we do.

Okay, that makes a lot of sense. So we are just about up on time. Where can people find you? Where can they find out more about workera? And then we'll wrap up.

People can connect with me on LinkedIn.

And also find Workera. We post content, we talk about our updates on a regular basis. And so I would encourage people to reach out directly on LinkedIn if, if we can help with anything or if, if they. They just want to connect.

Keon, it was a pleasure having you on. I love your perspective. I love, I love what you're doing organizationally and it's refreshing because a lot of what you're saying makes a ton of sense, just in terms of how much learning has changed, especially over the last three to four years, or let alone over the last four to 18 months on top of it. So thanks for being on the show and I was happy to have you, man.

Yeah, thanks, Ryan. It was great chatting.

On the next episode.