Name: Architects of AI: pivotal moments
Uploaded: 2025-04-24T16:47:10.817Z
Duration: 45 min 59 s
Description: Architects of AI: pivotal moments

Transcript for "Architects of AI: pivotal moments": Alright. Welcome everyone to our webinar. We've got pivotal moments talking about some women, who are architects of AI. So just as a a brief reminder before we get started, those of you who are just getting to know Cohere, Cohere is a a leading security focused, enterprise AI company. So all we do is we build, the entire vertical stack to make sure you get value out of, your AI strategy, everywhere from the foundational model all the way up to the the application itself. And just helping companies get to production is is our focus. So today, we to celebrate International Women's Month, we've gathered a few, a few women across our business who are helping create AI and help, helping us drive forward this this change as an industry. So, we'll do a few brief introductions and then, and then head right into to some questions to learn about everyone's stories. So I'll intro myself briefly. My name is Autumn Mulder. I run infrastructure and security here here at Cohere, and I've been here for about two and a half years. And I'll pass it along to Prutha first. Hello, everyone. I will start by saying that I'm delighted to be a part of this esteemed panel. I'm Prutha. I lead the security team, here at Cohere. I've been here for a year and a half. My professional journey has spanned multiple roles over the last couple of decades, working as a security evangelist at a big four to a threat researcher and ethical hacker in my early days, and then a product leader at large enterprise companies like Wallace and Palo Alto Networks. Outside of my professional life, I like to explore the outdoors and, also equally happy painting with my old pastels, indoors. So that's me. Thanks again for having me. I will popcorn it to Claire. Thanks, Prutha. My name is Claire. I've been with Cohere for the past four years, and I oversee the human data function here. So my team and my workforce is responsible for creating training data for these models, providing deep subject matter expertise. So we hire doctors. We hire mathletes. We hire Olympiads, that help kind of hone our models in different areas, as well as our team does things like running human eval. So, we partner with the modeling teams here at Cohere to help ensure that these models that we're bringing to market are truly some of the best that we have to offer. I will popcorn it over to Kelly. Okay. Thanks for having me. I'm Kelly. I lead the multilingual modeling team at Cohere, so we handle anything that's happening outside of English. Prior to joining Cohere about two years ago, I finished my PhD at Johns Hopkins University. Before that, I was an engineer at Google, for a handful of years. And, yeah, that's about it on my background. I will popcorn it to Tatiana. Thanks everyone for joining us today. My name is Tatiana. I lead one of our sales team. I've been with Cahir for almost one year, and I always say that I have the best job at Cohere because I have the privilege to work with our customers and potential clients and partner with them on the most sophisticated challenging AI problems and find meaningful solutions and drive impact to the business. Awesome. Thank you everybody for for sharing sharing kind of a brief background, and really excited to to talk more about your your stories. So we'll just kind of go around and hear hear a little bit more about, everyone in the group and how you've how you've gotten to to where you are and the problems you've been hitting over the the recent past. So first, I'd love to hear Claire. Love to hear a little bit more, about your your history, kind of your journey here. I know you've had quite a few roles within within Cohere. You've been a you've been a staple at the company. So I know you've done done things with marketing, head up marketing services, your product manager, you did chief of staff, and you just really helped this company grow. But your most recent challenge, as you pointed out, has been, how do we scale up the data annotation efforts within the within the business? So can you talk about your previous experiences and how those have contributed to doing this, like, massive scaling effort? Yeah. Absolutely. So I'll add a little bit of context because I think the story in itself of how we started this human data function at Cohere is is quite funny. And, at the time that we were just starting to explore human data a few years ago, Aidan, who's our founder, he came to me and asked me to find, you know, 20 people. We our office is down the street from the University of Toronto, and we thought that students would be a really good demographic to play around with our models, give us some early feedback. What did they like? What did they not like? And I kind of took that and ran with it. So the first thing that I really focused, in on while while building out this function or or seeing how humans and models can interact, was this idea of experimentation. And experimentation is a really familiar concept to me in marketing and in product management. I built a really strong muscle for experimentation. So if you're not familiar with those fields, things like leading customer interviews, running AB tests, that's the norm. And so in the early days at Cohere, I literally called in friends and family of our Toronto Office Workers. I had them join us for a three day, annotation pilot program in the office, and this helped us really hone in on the ideal candidate profile. So what does our target annotator look like? What kind of experience do they bring to the table? It helped us, you know, experiment with task design. So how are we actually teaching, newcomers about this software that we're building? How do we ask them to complete a task effectively or engage with these models effectively? And, really what these early days experiments really helped us do is figure out what's working. It took a lot of trial and error to figure out how people engage with this new type of technology. And iterating with a smaller group early in the process, I found was really, really valuable. It prevented us from making really costly mistakes when it comes to things like, you know, onboarding, annotators or when it comes to learning and development, how to teach them about this technology really effectively. It helped us identify, who we want to hire, who is, like, a a really successful and a great candidate for this role and who might be, you know, not not the best at kind of, fulfilling this type of function. And then once we'd really nailed down that early kind of MVP or that concept and we'd tested a couple things and figured out what worked best for us, that was the point at which Shin, you know, I called in real experts, our talent team, our recruitment team, to help build and solidify really, like, reliable and scalable hiring processes. So my advice here is really experiment with a small group of people. It allows you to make way more mistakes. It allows you to experiment at a much higher velocity. And then when you feel like you have that initial target, really nailed down, that's when you bring in experts to help you scale to the next hundred, next 200, next 400 folks. I love it. Love it. And it's it's so clear from everything you just described how the breadth of your experience has really helped you, make, you know, a great annotation function and, like, scale out that that entire organization. I I know I personally have seen, amazing things out of that group, and so, you know, huge kudos for for all you there. It's a huge challenge. Alright. Well, thank you, Kelly. I would love to hear a little bit from you, because you actually have, a a very interesting interesting background. And I I think you mentioned to your your history at Google, You were actually part of the acapella group there, which I had not heard until until we were prepping for this. So I'm gonna say it wrong. Googapella. Is that right? Googapella. Googapella. I love that. There's also alphabet. Oh, wow. Alright. The amazing things that you get out of some of these large companies, I just I love it. So, anyway, so you were you're in the acapella group, and you've talked about, both in your research career as well as what you're doing here, which is very, very research oriented, and the the kind of singing efforts. You've talked about how all of those are really, like, artistic endeavors. And so can you talk a little bit about how that that perspective has shaped your your research career? Yeah. So I think that, one thing we do in music, but especially in a capella, is, combine parts of things that already exist in an interesting way to make something new. And I think that perspective has, sorry. We're getting a little bit of feedback. I think that perspective has shaped my research style in which I think that I often take the position of taking bits and pieces from different academic areas and combining them into something that's interesting and takes on a new perspective. I think this partially comes from the fact that I came into engineering slightly later in life, so post college, which gave me a beginner's mindset. I wasn't so stuck in the traditional ways of, like, this is how things are done in computer science or applied math. And during my PhD, the thesis that I wrote was really a hybrid of natural language processing and applied mathematics. It brought together people from both departments in both academic departments at Johns Hopkins University who hadn't spoken much, and I ended up with a project that was focused on high dimensional geometry, but also machine learning, which was a unique perspective. To get a little bit nerdy on that, what we were thinking about is during our process of training word embeddings, I was, injecting measures of high dimensional geometry into the loss function so that when we were training the embeddings, we were thinking about what is the shape of the space, how does it relate to the one that we wanna use it with. And this was a pretty unique perspective. Love it. And it's kind of this theme that, I hear over and over about how innovation really happens when we bring together some various unique perspectives, and and bring that that, forward. Like, that's that's kind of what what most of our companies are are built on, right, is bring together a lot of unique perspectives and combine them in a way that you haven't seen before. So thank you. In fact, on that same topic, Prutha, I would love to talk to you about and hear from you about the the security angle and AI, which is a really interesting unique perspective and one that's just fundamentally critical for us, here here at Cohere. You lead you lead all our security, all our compliance programs, and that's a challenge to say to say the least in a rapidly evolving space, like like AI. So can you talk about what's attracted you to to this intersection of AI and security, and then what's exciting to you about about what's coming in the future? Yeah. Yeah. Absolutely. So, I left out one small detail in my introduction. I actually really faced a pivotal decision. When I moved, to this country, two years, two decades back, you know, I had the option either to specialize in robotics and AI or cybersecurity. It was almost a coin toss moment for me because I was truly conflicted, and I opted for cyber and spent the next fifteen years, building security products. My interest in AI never waned though, and I found myself drawn back to it at at which point I decided to focus specifically on on AI security. So as they say, life comes a full circle, and, here I am at the very intersection of the two fields that I was once conflicted, about choosing, between. What excites me most, I will say two things. First is definitely security for AI, which, like, Autumn, you mentioned, it is core to us here at Cohere and to our customer base. Our secure AI approach is very holistic. You know, it extends beyond traditional security. We think about it from the entire stack perspective all the way from the application, the infrastructure, the models, the agentic layers, the vast datasets. And, you know, thinking about this, like, very complex yet, you know, very unique problem is is absolutely, exciting. I am also very excited about AI for security, which is using AI to improve cybersecurity outcomes, whether it is automating tasks, identifying threats and secure code. We have internally seen, you know, operational efficiency gains, and there's definitely a transformative potential, for these security specific, use cases, which, I'm also entitled about. I love it. And I I I think you're absolutely right. Like, the, you know, this is so foundational, and there's so many pieces. And I I love the approach you've taken of just, you know, we don't slice up the the stack. We don't slice up the pieces and say, oh, we just focus security here. It's everywhere. It's up the up and down the whole stack. And I know that's been a huge, it's challenging, especially, because you don't want to to make make a function that's too large and slowing down the the company, and you've done an amazing job at kind of balancing those, those two while keeping us secure and keeping the the team involved in all the layers. So and I'm really excited to see where the the security function continues to grow here. So we we have a few people that are from engineering, and we've kind of talked about each of those things. And last year, I think we focused almost exclusively on engineering, but bringing AI to market is a team, team game. And so, Tatiana, we're really excited to to have you here to bring your sales perspective as well. The all of this is, is critical, and that's one of the the critical functions I'm excited to, be able to to collaborate and work with you on. So I would love to hear. You've, you've been part of the the president's club multiple times. You are, which which to me is an engineer. Like, what's a president's club? And learning about that has been has been, interesting and very cool to hear learn more about how do we, help go to market, how do we help companies, adopt adopt AI? So I'd love to hear from you. Like, how have you been able to to lead from the the sales side? What does that look like? How have you been able to overcome some of the what I imagine are quite rough patches and and difficult things, leading in sales, but especially leading in sales as a as a woman in the field. Yeah. Thanks so much, Autumn. And I'll probably give a couple of examples that, no one has years and years of experience leading or team and selling AI solutions. It's it's really new. Right? And, I think kind of, what I really encourage everyone, especially if you're trying to get into this field, really sit down and, figure it out. What is your skill set today? What drives you today? What is that gap in terms of the knowledge, in terms of the skill set you're trying to acquire, and how you can get there? And I would give an example, and I've been with Kozhikode for almost a year. And, I had a setback in my career that actually, required me, you know, to go and push myself outside of the comfortable zone and, find a new role. So I worked at a company for six years. It was a public company, and they were acquired by private equity. And, eventually, my job was impacted. I was promoted five times over the course of six years. And after the last promotion within three months, I got the news that my team will be eliminated and, my job will be eliminated. And I was offered the job that I had six years ago as an individual contributor. So in those moments, I think it's very easy to, start doubting yourself and start kind of like, oh, wait. I what could I have done differently? I think instead of that, I embrace the situation, and I really refocused and reshifted my mindset around, well, this is the time to go and pivot. And this is the time to go and really put the all the effort and energy and land the job of my dreams and asking myself, hey. What do you actually really wanna do, and where is my skills would really benefit? And I really wanted to join an AI company, and where I can build something from scratch, where I can help scale, where I could be embrace the challenge every single day and work through that. So what I really encourage people when you come across those situations is actually, push yourself outside of the comfortable zone. Get yourself outside of, experiences that you've never done. No one done AI. No one knew anything around how to actually scale a team and grow the team and set it for success. And I approached that process as a sales cycle. I looked at it as the funnel that in order for me to find my dream job, I can't just look at it. Okay. I'm gonna go and talk to one company. So I'm an equity focused on creating, a Google Doc with all the jobs available in the market where it will be the right fit. But instead of going and applying, I tracked down senior executives at each companies and, connected with them on LinkedIn and proactively was trying to get myself in front of them to have a business conversation, to have a point of view about their business, to be able to come in and actually, not, I would say, position as like, hey. Why am I great? But look at it as, like, hey. Where's based where you are in as a company, these are the things, you know, that I can bring to the table. And going through that process, I think it's, eventually, you know, I was able to, land on my dream job, right, and, join Cohere. That's been, I would say, a very challenging and rewarding experience. So the lessons, I learned through that, no matter how hard it is, I would always encourage everyone just keep continue pushing. But really in the end of the day, focus on what's important for you and where you wanna be in your career and set up, like, strong goals and execution. And no matter how many setbacks you have, right, a combination of small failures will eventually drive success. Love that. And and I think, like, the things you're saying resonate so well, because it's not just in sales. Right? It's everywhere, where you don't want to let, you know, small setbacks, you know, hold you back, and you need to just kind of be clear about what you need, and and move forward, right, in in your in your career. So Yes. Absolutely. And I think we're all right face all, face these situations in our life. So, Autumn, transitioning to you, you've been in financial services for three years. You worked for a very large nonprofit for fourteen years. How did you end up at Cahir? And looking at those diverse set of experiences, how do you think that help you to shape, right, and drive your success of where you are today? No. That's a that's a great question. And it kind of as we're focusing on this idea of, like, pivotal moments and and, career transitions, it's, you know, that that was definitely a big one for me. I was very explicit, to kind of reflecting what what you wanted, and how you how you found your way here. When I was, I've been at the nonprofit for fourteen years. I've done that. I actually I have three kids, so I was part time for about seven years with that nonprofit. And, it was a great experience, very, very large company that, you know, was kind of like supporting 60 different businesses. I worked in the data field. I worked in infrastructure. It was just a really great opportunity. But, then I transitioned to a public facing company for three years and enjoyed that, ran infrastructure, did a lot with DevOps and internal, systems. It actually supported a pretty large and extensive, machine learning organization. But I knew in 2022, I knew that what I wanted to do was work at, a startup and specifically one that was in tech doing something interesting. I didn't particularly care what kind of startup or what kind of tech. I just wanted to to have that that experience, for myself. So I went around and started applying to to a lot of different, different startups. And it was really just serendipitous that I that I came across Cohere, talked with some of the people here. At the time, that was before, kind of this AI revolution. Not any and nobody had heard of ChatGPT because it hadn't it didn't exist. And so I I interviewed at the company and said, oh, yeah. You're doing this thing called large language models. What's that? That's that sounds really interesting, very cool. There's no market here. So I'm just gonna join this company and go down with the down with the ship. So it's been a really lovely thing to see that, we as a as an industry have been able to kind of turn that corner and say, hey. Look at this really cool technology, and there's actually clear business value. We know, how it can be useful. And being part of this effort to figure it out and build the plane as we're we're flying together, has been really enjoyable. But, that's where I'd say, like, all my experience from working in nonprofits, working in public but financial services, companies that were not tech focused, that's where I bring a lot of that that, experience to the table here here at Coherent and say, look. Our customers are generally non tech companies. We're helping you know, sometimes we have we have tech companies, but but often we have a lot of these these companies that say, look. We don't build tech. We use it. We really want to understand what's the easiest way to to get the latest and greatest into our environments, but do it in a secure way. Do it in a stable way. Do it in a way that meets the needs of our our customers. And that's, like, all the experience that I I had previously is what's kind of funneling and shaping the the vision of how we how we build products like like North and bring those into, customer environments. So I love that. This kind of hearkens back to that intersection that I think all of us are bringing to the the business and our and our jobs. So let's actually flip. I'd love to hear a little bit more again. So, Claire, I'd love to head back back around your way and hear about looking at your diverse roles and looking at kind of the the intersection of things you've done. For you, what have been the skills and experiences that have been most most transferable? You know, how's that intersection working for you? Yeah. This is, one of my favorite questions to answer because, you know, throughout my career trajectory, I've, you know, really cut my teeth at startups. One of the things that I love doing is that kind of early day startup process where you're wearing multiple hats, because it gives me an opportunity to find new ways to apply and really stretch my skill set. But part of kind of my unique selling prop, and part of what I think makes someone really valuable in early startups is the ability to be scrappy and resourceful. Again, you're working in a resource constrained environment, trying to get shit done, and sometimes you have to think really creatively and out of the box. Examples of what that's looked like for me at Cohere. You know, when we stood up our early days of an annotation program, typically, you would, you know, send out very legit contracts and have a payroll system to be able to compensate people for the work that they're doing on the project. We did things like sending out a Google form, getting people's PayPal emails. And when they'd spend, you know, five hours in the office with us, we would just go into PayPal after that shift and and send out payments individually. We didn't wait until we had these kind of perfect systems set up or perfect tooling set up before we just, you know, dive right in and started doing the thing. Another example that I think of really fondly is, just this idea of how do we find our first fifty annotator hires. We wanted to do it within a couple weeks. We'd never done it before. What our team did is we showed up to every university hackathon within the Greater Toronto Area. We would go and we'd hold group interviews on the spot, where people would come in. They'd spend time with us, and they could walk away with a physical offer letter in hand knowing that they had a job after that experience. When we ran out of hackathons to show up at, we repeated kind of that same group interview process at our Toronto office. So if you're, you know, an employee of the office at the time, it looked really funny. We'd have hoards of folks just show up, for three different on-site interview sessions. At the time, it was just myself and kind of an HR and talent contractor that we had that was doing this hiring funnel. And, like, you know, we couldn't be in three different places at the same time. So what we did is we had our first few hires actually help us hire the next 50 people. We each ran a different room, a different interview session, and, it just shows how much you can get done when you, have few people, but you're thinking a little bit creatively. And, again, like, these things aren't scalable now. That's not the way that we're hiring today. But I think of them as really vital and instrumental into our early successes, being able to prove out this idea of human data at Cochlear to justify getting additional investment, additional hires to really make this function a little bit more legit. And so, again, I think my advice for folks, when they're thinking about their skill set and especially when it pertains to startups is, think about how you can be resourceful, how you can be scrappy. And when someone comes to you with a really ambitious ask, lean in. Let that be your differentiator. As Tatiana said earlier, I think, always think of ways that you can push yourself. And for me, I think my learning here is, don't let perfect be the enemy of good. So, dive in. Try doing it. You know, you've got time to make things better, and decide on what kind of your final form looks like. But I think a lot of experimentation and resourcefulness and creativity in the early stages of the process, really kind of, is what led us to, the program that we have today that that I think I'm I'm really proud of. And it's also like you know, those are the really, really funny stories that I think of, and that I can share with folks that are joining the team today, that give an give them an idea of what the culture on the team looks like, what the culture at Cohere looks like, and what we'll do to to get the job done. I love it. Yeah. Thank you. Thank you, Claire. And, everything you're saying about the kind of mental fortitude and the ability to to quickly shift between things definitely definitely resonates. In fact, I think, Kelly, you've you've probably got some thoughts about that as well since you were a founding member of the the multilingual team, probably have a lot of battle scars of how you need to to have that that fortitude. So what's your what's your mental framework for dealing with ambiguity? Yeah. Sure. And I'll just give an example of, you know, what happened during my first, six months to a year at Cohere where I needed to really wade through this ambiguity and figure out how to get things done, despite that. So basically, when I joined Cohere, we were, you know, primarily English only. I knew that I needed to if I wanted to make the model work multilingually and it was just me working on this, that multilinguality has to be injected everywhere. It's not the kind of thing that you can just sprinkle it on like another capability at the end. So I spent the first few months really understanding the entire tech stack all the way from our tokenizer, pre training, post training, reward model training, preference learning, worked on the first multilingual evaluation, and even did it data annotation. So my first six months were spent with about three to four weeks on each part of these pipelines, ensuring that we could use these in a multilingual scenario. One of the big challenges there was a lot of this stack was set up specifically for the English model. So some of this was even working with the folks on serving, when it turned out we couldn't serve a multilingual model because it was a slightly different architecture. So working through all of these issues, and it all, it all kinda came to a head in December 2023 when I remember right before the holidays, we had a big group meeting, and I was able to say that the entire pipeline is now integrated. These are our scores on a handful of languages, and it was a proof of concept that we could train a multilingual model at Cohere. And then the next step was just how do we make it actually a good model. And the rest is history because now we do a lot of multilingual training. We do. And and I love that because the the reality is that's become one of our one of our strengths as as the business. So, you know, all of your your efforts in kind of pushing that forward have become, really, really critical and and, incredibly valued at the at the business. You know, pivoting to to kind of another area that's also incredibly valuable and very, very pivotal for the the business, proof that security is, like many many industries, but particularly noticed in security, it's it's it's a challenge. Right? Because you're you're all about risk reduction. But as we just discussed, there's a lot of ambiguity in this this space. People are figuring out what are their standards and then to say, how do we make sure those standards are secure? You know, there's just a probably a lot of a lot of pushback. And sometimes this this push and pull of how do we make sure it's the right level of security, well well moving at the the right speed. So can you talk about how you've navigated resistance both from honestly, both both sides. Probably the, securities, the engineers that are advocating for that and then the the engineers that are advocating for speed. So what enables you to influence those decision making processes? Yeah. I will say that this question has not aged, through technological advances. In fact, it gets even more relevant with time. So I I truly believe, that, you know, security, like many other functions, is is a business risk. Alignment on a risk management framework, where your stakeholders, whether it's the engineers, you know, cross functional teams, anyone, where your stakeholders are bought in, I think it's key to advocating, for for some of these measures. Again, the risk profile, as you mentioned, you know, it can look different for companies. Some are more risk averse, some are more tolerant, but buying is is important, to establish, that as, security as an integral part of your ecosystem. As far as navigating resistance goes, I think trying to understand the root cause, and meeting engineers where they are, allows you to strike the balance. But I also have to be, honest here. It doesn't always work. No. Rome wasn't built in a day. I I will say that, right, patience and persistence, are really what it takes to establish security as an enabler, and that's what we try to do every single hour, every single day. It's a good day if we're able to write meet someone and help someone, you know, that that particular day because, we've tried to, sort of, right, do that, like many other teams and help, help those teams. So I think that's, that's it. But sometimes, you know, it just takes, takes time and then repeatedly, you know, trying to make that that same message and, share that with the stakeholders. Yeah. No. And I love what you say about, like, look. We can all be honest. Sometimes we all strive for this and we hit here. But that that kind of proponent, or element of persistence is so critical as as a leader. I know that's something I've heard repeatedly from from the org, as people are interacting with you and the the security org. Just this idea that we aim here. Even when we hit here, we keep we keep pushing. You know, Tatiana, I actually think that's very relevant from a from a sales perspective as well. And I know you're a huge proponent of the idea of grit and persistence. So what is what does that mean to you, and how do you become gritty or persistent? Thanks, Autumn. I think, obviously, great AI in sales is probably number one skill that you need to have. But in a startup world, it's applicable to annual. The way how I define great, it's a combination of determination, hard work, perseverance, patience, and internal drive. An example I would like to share, and I think that probably would be relevant to a lot of people, when you look at someone's LinkedIn profile, you see the success. And you think that the career is very linear and you think, oh, wow. This person was lucky or this person did all these things and they achieved all these amazing things. But what we don't actually see what's behind. And a a lot of that success comes with a lot of sacrifices that those people have to do as well as with lots of setbacks, and any career is not linear. And if anyone looks at my LinkedIn or my resume, you'll see that, I've been in sales, and I've been number one individual performer and moved into leadership many years ago, and led successful teams. But what people don't see is that it actually took me ten years to get my bachelor's degree between two countries and three universities. It took me six months to find that first job out of college when everyone was closing the doors and it was almost impossible to get any job. People don't see that I was rejected from Salesforce three times before I actually was able to land a job. Right? And that I had to take every possible job from waiting tables to work in a nonprofit and being a paralegal and going to college at the same time to support myself. So those experiences, I think that's what shape us. Right? And I think that's what Greek means to me. So I and a lot of times right now being, a people leader, I look at it. There is a lot of and I think the questions was asked. AI is constantly evolving. Right? What skill set can I focus on? And I would say, look. Like, there is a wealth of knowledge. Right? The amount of information that exists today never existed more. Right? So you can acquire any type of AI skill set, whether you're looking for a technical role, whether you're actually looking for more on a go to market team. But what I really encourage people to I feel really understanding what it takes to be in a start up and focus on the skill set around how hard can you push yourself and how hard can you run for this company and how hard can you work because that skill set is not coachable. That's a skill set that we're all being in a leadership role looking for in people because we can teach you something transferable. You can learn on the job, but those are the parts we expect everyone to bring to the table. Love it. And, and I agree. Like, this is this is a a kind of broad and evolving field. And, like, at the end of the day, I think anyone is capable of learning those skill sets. Just don't get discouraged. Right? And stay persistent, and keep keep trying. Right? You'll you'll get there. Yeah. Absolutely. And, Autumn, I wanted to transition to you. You've been at Cahir for a while. Your job was very, very different even a year ago comparing where we are today as a company and how fast the industry moves. And the change, I think, is a DNA of being in a in a company, like Cohere. In your role, right, managing infrastructure and serving models was very different comparing where we are today moving into the application layer and working with all the complexities to serve those applications securely and being able to connect to data sources. So it would be helpful to understand how do you adjust to sales, right, and how to the change, and how do you, help your team to operate in a such fast and dynamic environment? Yeah. I think these are excellent questions because, there isn't you know, kinda like like you said, Prusa, there isn't, like, it it ages well. These are think, things that we see no matter the no matter the time. So from my perspective in yes. Infrastructure is evolving pretty rapidly. So is security. So so are most of these, most of these fields. And so the the answer is similar. From a leadership perspective, I think it all comes down to surround yourself with with people you can trust, surround yourself with with a team. Transferring information between human brains is the slowest form of communication. And so what's most important is that you you get a team in place, you and and your team, those people that you are collaborating with, you all understand, kind of intuitively what what you're thinking about, and that just takes time. And so it's so important. If you're going to manage this pace of change and the speed, it's so important to to really be human focused first and really be focused on the the people you're working with. Make sure that they have the information in the context you're operating on and that they have the the information, about what's coming, because there's no way to just tell everybody everything. You guys you have to make sure that the people you're working with are making wise decisions. But even then, like, things will things will fall apart, and you just kind of have to do something, at the the the last minute, if if the situation arises. So I'm thinking, like, a month and a half ago or something, we had an infrastructure situation where something had changed, of course, because everything is fine if nothing changes. But something had changed, we'd, dropped a a certificate in production or something. And so, we had a situation where for whatever reason, our on call situation didn't actually alert the right person, but I ended up getting getting a ping in the middle of the night. And the people who could could probably resolve the problem the best weren't there, just just due to to time zones and the way things were set up. So, you know, we're kind of in the middle of the night. There was somebody across the world who happened to to be up at the right time and was able to walk me through, some of the changes we need to make to to restore the certificate in production. And so, like, being able to manage these, you know, this constant change, it's it's kind of a a mix of make sure you have the right types of skills and you're not cutting yourself off to any anything, if I hadn't stayed involved with the from a technical level with the systems we have and the ways that you need to go in and do something like restore certificate, I wouldn't have been able to do that for the team. But there's also this level of trust, you know, in the individual that that we had who was able to walk me through that situation. If we didn't have that, like, level of synchronization, if we weren't able to just quickly via Slack, say, oh, here's the problem. Go look at this, and resolve the issue like that. That also would have been would have been a challenge. So it's really those those two things. Stay stay aware, stay stay involved, keep learning, and make sure you trust the team that's that's around you. And, with those two things, honestly, you can you can hit any challenge. So I think we have five minutes or so to to answer a few questions. So I think we had two that I saw in the in the channel. And if anybody else has one, please drop them in, and we'll be happy to to answer. So first, with the we I'm gonna ask Kelly, actually, because you're on the research side. So with the field of AI changing so quickly, how can you develop the right skills, and how do you navigate these constant pivots? Yeah. So one thing that I try to think about is what skills are for now in ephemeral and what is, like, like, a core skill that will be broadly applicable. For instance, my most explicit advice is study math. If you have time like, if you're still in academia and you have time to do it study math, it will be applicable to anything. As soon as your technical field changes, you'll be able to read papers in the new fields. If you have a strong math background, reading deep learning papers can sometimes be a breeze. And the other thing is, be a strong engineer. I know that AI tools are really helping us code, but you need to write clean code. There was a if you remember from, like, Twitter or x a couple years ago, it seemed like some people thought that prompt engineering was be gonna become a job. This was an ephemeral thing, and that's what I mean by, like, pick the things that are cross disciplinary, don't change, such as I'm gonna plug math and strong software engineering skills. I love that. I love the math plug. Never could do that myself, but I actually if you can, it's a it's an excellent field. I'll ask the second question. So this one's for you directly, Claire. So you mentioned your unique value profit and recognizing that. And, how would you advise someone discovers their own unique value prop? Yeah. This is, one of my favorite questions to answer. I think there is really two components of it, when I think through how to define what you bring to the table and what sets you apart. I think the first piece is really starting with self reflection. So, you know, take a peep piece of paper, write it down, think through, like, what are your most significant accomplishments, what skills can you personally identify, that contributed to those accomplishments or those achievements, And then think through, like, what unique life experiences do you have that help lead to those accomplishments or that other people just don't have. So examples of what that could look like might be you have expertise in a niche area or you've got a really specialized skill set. It could be that, you know, you think you have really strong leadership abilities, really strong team building abilities. Maybe you're really great at establishing or building consensus. And then the second thing that I think is really critical to think about and think about often, is what do your peers and and the folks that work with you or interact with you, really have to say about you? So one of my favorite interview questions to ask is, and I and I think it gives people, like, a a great opportunity to humble brag a little bit, it's just if I went to your coworkers and asked them, you know, what do they turn to you for, how would they answer? Usually, that'll take you pretty close to how, you know, people, envision you and what people identify as your strengths. I think when you reflect on those specific anecdotes, it really helps you uncover what your best skills are and what your impact is and how people perceive that impact. So if you are just starting your career, again, you might not have coworkers that you can ask this question to. Think about, you know, asking this question to your friends, your classmates, or your family members. Again, it doesn't have to be limited to your working experience, but you'll often get, you know, really strong and great anecdotes about what sets you apart. Maybe you are a person that always gets things across the line. You you're, you know, that that linchpin in in group projects at school. Really, I think spend a lot of time thinking about what sets you apart. See if, you know, your impression of what sets you apart ladders up or matches with, others' impressions of you, and then find a way to create a really strong narrative and and come up with really great examples that you can share with prospective employers or you can share with, you know, your boss when you're asking to expand your scope or your reach. I think those are kind of the two dimensions that I would think at. And and I love how there's, you know, threaded through this all. There's kind of a sales element almost of understanding yourself and what, you know, how you're positioning yourself relative to to people who need your unique value. Alright. We got a lot of questions, and I love I love these. So I we we probably only have time for one more, so I will pick, this one. This one's actually for you again, Kelly. How do you strike the balance between actually hands on keyboard, writing code yourself, and, the the vibe coding? Yeah. I I don't know. Opinions, but I really wanna hear you guys. I think this was easier for some of us that came in, like, a little bit earlier, because there was no AI coding when I was learning to code, how to do everything myself. I'm still a VIM coder to this day because I love it, and it makes my workflow faster. But I'd say just make sure that you can do it yourself first. And if you can, then, of course, use Geni AI tools, but it shouldn't be a crutch. So maybe if you're working on academic assignments, do them manually. Make sure you know how to do it. And then if you're, you know, coding for your job and it's allowable, then certainly use a AI tool to make it faster. I certainly do it. For instance, I I'm not that great at Pandas, but, you know, AI tools are. So that's where I will typically use it when I don't know the syntax of a language, but I know what I wanna do. I love it. And and, again, just from a security and infrastructure perspective, certainly, we've got strong opinions about VIVE coding because I've down. You know? You have to understand the system that you're designing for, and so I love your your advice around you know, it's almost like learning math. If you don't know how to do one plus one yet, maybe learn that, and then you can start using the calculator. Maybe, you know, learn the fundamentals of of algebra, and and you'll you'll get there. Well, I believe we are at time, and I know there were some some lovely questions. We appreciate everybody, joining in and listening to us for this this this really enjoyable conversation. It's such a such a pleasure to work with a a group of amazing women here at the the business. So thank you. We'll see you all soon. Thanks, 701.