Episode 82

Qiang Meng on Balancing Data and Intuition in the Hiring Process

TIM: We are live on the Objective Hiring Show with Qiang. Welcome. Thank you so much for joining us.

QIANG: Not at all. It's my pleasure, Tim.

TIM: It is absolutely our pleasure. And for the viewers who are listening, not watching, Kian's got like the most ridiculously beautiful background with all these wonderful plants and decoration. It almost looks fake. That's how beautiful it looks. So Thank you for giving me like a Zen feeling when I'm looking at you.

QIANG: Pleasure again.

TIM: I'd love to start our conversation today with a discussion around a bit of a yin and yang feeling that I always have in hiring, which is, you know, when to use data, when to use your gut. And I personally kind of think of this as like a spectrum and. Maybe people sit on different bits of the spectrum in terms of how much they rely on one or the other. Have you thought about this? How do you balance the sort of intuition versus using data to make hiring decisions? Yeah, I'd love to get your, your thoughts on this topic.

QIANG: Yeah. I think this is a really good question. First of all, from my personal experience during the recruitment of past 15 years, I personally feel there's no way we can hundred percent like find the best candidates based only on the objectives opinions and so on and so forth. Like what you mentioned data, we can try our best to use that part of like knowledge to find the best candidates such as For example in my interviews to find the best data engineers no matter me or my team. We all often have an interviewer and a shadow. so in the we make sure like The shadow often goes to the interview but mute themselves and close the camera, but they just listen. Once the interview finished, basically we have a quick discussion from both of us, interviewer and shadow, to make sure like our opinion or decision making is really like as less bias as possible because human beings always have a bias. A second thing we're trying to reduce is kind of like a bias or like a an intuitional subjective things. We, we list, actually, we try to have a list of checkpoints, that's including the technique points, but also the soft skill points. And each of the points we give a weight. Like for this position, this skills is how much more important than that skills and so forth. So in the end, we try to have a number to representing as fair as possible to the candidates, technically speaking, soft skill speaking. But all in all, there's no way to have 100% objective opinions in my personal opinion, but also at the same time, I don't think it's a good interview process. If the decision is made a hundred percent by data, I would say it's more like a reasonable part of percentage from between the objectives and the subjectives. Intuition like your feelings is more like 80 20 principle again 80 80 percent of that I would say data really matters. We need to have a framework We need to have a team alignment about what is so important for this position this candidate to be really focused and get ready to start the job, 20 is still there to see the feelings and The culture fits especially when you need an engineer, it's not like a data engineer, but maybe a lead data engineer or data engineer manager, then you deal with people need to engage with stakeholders how this person's way of working. What's their mindset? Does this a person fits into culture is something very, very important. And if you ask me to that a priority, even technical part is a data part is 80%. 20 percent is like the feelings, but the feelings 20 percent is more important. Then the, the, the, the data part, the technical part, why I'm saying that, because often, you know, especially in the data or AI domain, the technology revolution goes so fast. you remember to check GBT or generating AI, it's only appears for one year or two year and how much they already changed our lives. So therefore, method, the way we're working, if these candidates. He's a fast learner. Does he always keep tracking of the latest things in the market and keep learning and also teaching and coaching his or her teammates. It's much more important than for the moment, if the candidate knows about GCP or Gen AI or something like that. So that's my opinion. And One last thing I want to mention here for this question is like very important often for any candidates You go to apply a job, but the job didn't give you offer. Don't give up. It that means like you are not good all about like if this person this candidate is really good fits 100 percent into the position That as I mentioned 80 percent like the real fits, but 20 percent is like the feelings the culture so so forth I hope I get, yeah,

TIM: Yeah, that's a great, a great and comprehensive summary. One thing I've been playing around with is I feel like, Even what you've described there, the remaining, let's say 20 percent intuitive gut feel feelings component. I feel like we could still make that more objective than it is. And we could still ultimately measure it, I think. So for example, let's just take cultural fit as an example. Most companies would evaluate that in some kind of behavioral interview where they're asking them questions and they're expecting the candidate to demonstrate a history of a certain value that they're looking for. So for example, in our company, We look for candidates who can make it happen. Candidates who put users first candidates who are no bullshit, like direct to people. That's what we're looking for. And so we would ask them questions to see if they can demonstrate that value by coming up with some kind of example, I feel like. With that approach, you could still make it measurable because you could say, well, I don't know, here's these four values we're looking for. Here's two interview questions per value using a similar weighting system to what you described before. So that at the end of the day, everything still has a number, even if it's subjective, it's like less than just making it up. What do you think of that approach? Is that perhaps it's actually what you, you do anyway?

QIANG: I think that's really valid points and I, I kind of agree even more because in our interview process, we often also invite our candidate to share with us their, their colors. I don't know if you know, there's professional company do the, the color test for people in the corporation lives, like which color you are. And one thing to be clear is like, we are not. Only looking for one color set is saying like this color is not better than that one We are looking for always to build a team. We all have a different color so that we are one team This is I really agree with you. But in my opinions like I was proposed saying, like, it's really hard to be 100 percent objective, and it's not necessary to be that. It's because interview process often has very limited time. And my understanding is, like, to understand a people, a mindset, a potential way of working, it takes time. You know, something like by a few questions, or report, or talk to 30 minutes, one hour, that's the conclusion you get. Really correct from what you are looking for. It's something very similar Like sometimes you see some candidates engineer CV is so beautiful But after the candidates on boarded for the job for the first one month or three months You realize like it's just too much filter in the beginning, right? but it's not really like something in one hour or 30 minutes, you can really Understand deep dive over there. So therefore the process need take time need to be comprehensive. So go back to your questions. I agree So we should try as much as possible to be objective To be like able to have some like fair process with numbers and weights and checkpoints to make sure like more or less they are there but we are not, in my personal opinion still, we are not supposed to looking for 100 percent objective solution for that because one of the very important point, we are human beings. we are not an object. It's not fair to evaluate anybody as a human being that By the numbers by the points by the structure, something like that.

TIM: Yeah, I think you've nailed a couple of what I feel like almost intractable problems in hiring. One is the small sample size problem. As you say, you've got like an hour to interview someone. Maybe you have three interviews with them and which is already, I'd say a reasonably lengthy process for an individual contributor. I know some companies have like seven, eight, nine, which is ridiculous, but even three interviews hour each, it's still only three hours with a human. And so there's obviously a finite amount you can learn from those three hour long interactions. And then the other thing I was going to mention was, at the start you were describing having a couple of interviewers. So you mentioned having, let's say, yourself as an interviewer and then a shadow interviewer, which is not a concept I'd heard previously. Can you tell me why you have the shadow interviewer, why they're kind of in the background? Is it because they can just focus on what the candidate's saying? That kind of makes it easier to to have a valuation. Like what? Why do you set it up that way? I'm interested.

QIANG: Okay. So let me get back to your question this way. So first of all let me give you a quick introduction about our recruitment process. So we basically have three or four steps or stages for hiring engineer Okay. Of course, first of all, we start with our HR team to have a basic filter and match, ask some questions about very fundamental things. But then we invite our candidates for a taking home coding challenge. After that, for candidates who survived for the coding challenge, we will have a quick Q& A. Technique fitting interview, but this technique fitting interview is not just ask like do you know about python? Do you know about spark and so on so forth? It's really follow up the coding challenge to ask deep dive questions with 80 percent of technique checks But 20 percent very important soft skills. Are they are motive? Are they really motivated for the job? Are they we're working like entrepreneur so that they keep driving by themselves self driving? They do the automation so so forth and last but not least It's only for some roles, very important for some management positions. We will have a leadership meeting, a leadership interview for the last round, but most of the time it's HR taking home coding challenge and a technical fitting for one hour. That's, that's it. So we try to keep the process as simple as possible as because as far as you know, data engineers are really like a hot. Market job precision is often very hard to find a good data engineer, so we don't want to make the process too complicated to scare the candidates away. Okay, go back to your questions. Why we are doing these shadows for interview? The major points is like everybody's different and everybody have their preferred method of way of working and also not so method of way of working. But me as a team leader often to try to Influence my team to saying like we don't need everybody the same we need actually have different colors Different people a different way of working into one team So so that for us may be perfect to be a data engineer or a lead data engineer but when we Sit together as one team and we are perfect, a strong team. that might be also the reason my first team in a, a, a U. S. then and company who has a more than 170 history was recognized by DataEQ organization this year as the top 10 global. the data and analytics team. So the point of that to have really like one interview were to try to be unbiased as less as as much as possible unbiased to ask questions. We agreed as team framework, but also the other side is really like a, a mirror of themselves once we finished this interview to connect to each other and also see if there's any gaps of understanding for this precision so that we can. Fulfill the gap to find the best candidate. And plus, one thing I want to mention, for this technique fit interview, we always ask the candidate if they mind we recording the session. Because sometimes, you know, even you have two persons for interview, verse for the interview, but they can have a really strong, different opinions to each other. So therefore, in that case we have the recording session. If that happens, It doesn't happen often, but if really that happens, we can share this interview with the third party, with the rest of the team to look and then have a fair discussion to say like, okay, what do we think what is the strengths of the candidate? What is the, maybe not so perfect for the candidate, but is that really matters or actually it doesn't matter. We, we are confident that we can coaching new training, the candidate to be ready in a very limited time for that position. So I hope I answered the questions here.

TIM: You certainly did. And straight away, I was thinking of an interesting AI use case there. I'm not sure if you've tried this or you consider trying it. Of course, privacy concerns permitting and whatnot, which is almost like an interview assistant. So You know, you've got the video recording tool, it's got the transcription, it's very accurate these days, feeding that into a chatGPT asking, Hey, like, these are the things we were looking to evaluate the candidate on. How did you score the candidate and almost having that as a, another neutral and for the comments, third party do you think that would work or would that help in any way?

QIANG: So far in our, okay, there's some context I need to share with you, first of all. So I think this AI thing so far, there's two factors. They help us as like a boost of the interview process. One factor they help us to. Smooth, Smooth operations, often like a schedule meeting you know, companies can also develop a chat robots for candidates, ask some basic quick questions about what's your team, what's your company's culture, what's the project that you're doing instead of like repeatedly, you answer the questions to, to, to the candidates, so on and so forth on the operating set. on the other hand, I don't really feel like they really can help us for the moment. To, to find the best candidate because me, my expectation for the AI tools in the future could boost the further for the recruitment process, especially for the technical positions. if they can really do some predictions of these candidates we're working can really better read their math, math sets from their limit is a precision of experience can help us to prediction their potential. Like how this candidate's problem solving is he or she is a keep fast learner, so on and so forth. That is really like the pay point that the AM can, can help my team, our HR team to solve the problem. I really appreciate now like the AI can help us for operations like Finish interview. We can have a quick real time feedback shares, some other things a schedule of course, but More important things I see AI still know there Maybe some companies already doing that if anybody knows feel free to ping me on linkedin and tell me the truth already have some really good AIs to predict their potentials and read their math stats and So on so forth. but so far I see AI is helpful but it's not like a reach to some expectations that can really solve the root pay point for the recruitment process for now.

TIM: Yeah, that's interesting around the kind of prediction. I'm. I'm assuming the gap here, maybe as it always is, will be the data because the data you would need to make those predictions maybe just isn't freely available. Like if you need, and this is, as you were saying before, part of the problem with the hiring process is it's a game of trade offs. Like you can't have 20 interviews and 50 tests or whatever, like you have to make your mind up quickly. Otherwise the candidates are going to drop off. It's probably going to be similar here, which is. You would need to collect, I don't know, comprehensively, their IQ, their personality, ask them all these different questions about their history and their jobs and test their skills. Like you would need to collect, I guess, a lot of data to make that decision. Maybe that's the hard bit as opposed to the prediction itself. Like maybe the gap is the data. What do you think?

QIANG: Well, I would not fully agree with your opinions about like people don't have enough data. if we talk about virtual learning in past a few years, so deep learning that. Indeed, often to have a good prediction models often you need really good quality of data and also a huge amount of data. not the case anymore. Now, like in the, especially with the help of generative AI, actually what a good model needs Is a good data. Even the data can be so huge amount of amount because the AI can think is most important in the data's, the logic and the relationship. So let me give you one example. Even with develop ai, when I look at the good candidates of data engineers, often I ask our HR team to provide me with their LinkedIn profile because why? I still need LinkedIn profile, even I already have a very comprehensive. CV of their past experience and projects because I try to find relationships. In the past such as I always look at their linkedin profile is their references and recommendations Who gave them references even the references can be very limited. What's their relationship? Does this candidate directly reporting to him or her as a line manager or this references is from the team? They're working together, right? this is very limited information data, but help me to With the link of the logic helping to really understand how this person's working in that job that's he or she is the team player with their team. So that's he or she is a good employee to have a good relationship. We are line managers are different line managers in different jobs at different time windows, so on and so forth. So go back to your questions, I think. Of course, in ideal cases, like we have a huge amount of data so that our model can be enhancing, can be training, but most important things my personal experience and opinions is not really depends on the amount of data, but the data quality is a data logical relationship and from what I know, The LinkedIn profile with one page of like really minimum amount of data can already have a huge amount of information by data to tell us like their potential, their way of thinking, their way of working, and so on so forth.

TIM: I feel like maybe the other, let's call it, Game changer, although that's been a horribly overused term that the like big unlock would be automation from where I could sit at least I'm, I think now the large language models are at a level where we're just waiting six months, 12 months for the application layer to be built on top of these LLMs to, I think, automate 95 percent of the process, like the sourcing, the outreach, the matching, the initial screening interviews, I think in a year, it would be considered normal that you would just be doing like the final interview because everything before that has already been done. If that were the case, hypothetically, would that be a big problem solved for you? If you just come in, yep, final interview, hour, meet the person, I already have their match score is like 97%. I already know his, their strengths and weaknesses. I'm just going to do like a final sanity check. If that were the case, would that be helpful to you?

QIANG: I actually kind of agree even more the time when you mentioned about everything changed so fast with gen AI, AI applications. I believe like in half a year, one year window, definitely speaking, like we're going to have some tools to help us there, but I still feel. I still feel like the interview process is not a good way to have 100 percent like AI driven decision making even in the final, because there's a bunch of problems like right now in my head. For example, the first one, the point like in the end is to find the best candidate, the best person who fits into the job. Because when you use AI to optimize the process, the candidate can also use the AI to against your process as well. You never know. Imagine like now we have so much like a future on such a beautiful CVS how could you prevent like the candidate don't know even better how to manipulate the AI to make sure the process is amazing. And then, you know, it's going to be like a not fair process for the rest of Canada who don't know how to do this process. Second, the same. This still have bias. I remember like in the prior use cases when we're doing the AI predictions in the retail business. One of the very major topics like how to reduce the bias as much as possible in the model itself. It's very tricky. It's very tricky, very hard. You need to continuous optimize in your model to reduce the bias and that's gonna be our second topic about bias. applications, how could we guarantee the bias is not there for the process? the third problem is still about a human being. Like I just personally still don't feel, is that a good things to 100 percent like process candidates in this AI way without like you really personally have a conversation, have a talk, have the feeling shares. Because this is human being process. It's not like you go to buy a garment buy clothes and so AI tell you like what is the best color fits for you or what the sizing fits for you But I always tell like my companies people who are doing the garments like a fabric have a soul So that's why human being like buy the garment still made by hand Rather than like made a fast fashion by the machines who is a cold metal so on so forth, right? And the last but not least is the still like the feelings matters. I don't know Maybe engineers is not a perfect domain to to marketing like as a failing matters principle But the engineers one day goes to a manager management positions or like a leadership positions I think it really matters you these people and his or her line manager has a really great atmosphere To work together It's so matters at least for people like me to looking for our next positions Like who's my line manager what he or she is doing what he or she's way of working I don't know if you see but in my first observations, I often see like A line manager, a people, they're all great. They all did such a great job in their person. You know cooperation lives, but when you put them working together, it just doesn't work Like there's nothing delivered. It doesn't work. And then why I'm still saying like, okay, yeah. So AI definitely can help, definitely going to boost 80 percent of our productivity. That's probably already great, but I still see there's so many problems. Like if we goes one day goes that way to see like, how could we further solve all the concerns and problems to make sure. The candidates don't use AI to abuse the process as well. Reduce the bias, respect the people and make sure like you really find the right person, they fits. The company, the precision, the task of that.

TIM: And you mentioned in passing in your process that you do. a take home project. I'm wondering what your thoughts are then on candidates using AI to do bits of that project. Is that something you expect? Encourage? Discourage? Have you written questions in a way that makes it hard? Like, what are your thoughts there?

QIANG: That's a really good questions. My, our team's coding challenge works really well until the gen AI appears and then give us a huge impact of that. However, we still having this process over there. And the questions in the process is still the same. You know why? So it's more like 50 50. 50 percent when they recruiting the candidate, as we mentioned, like it's not really just technical skills. It's about the mindset, the way we're working, right? Entrepreneur, we're working, keep learning fast. Because reason why we do taking home coding challenge instead of like a, you know, time scraping, like, you know, ask questions and ask people in the limited time is we want to find out the candidates. First of all, One of the common feedbacks when we hire somebody after the coding challenge, we're looking forward to hear like this candidate opens the camera, joins the interview, saying like, I had so much fun with this coding challenge. You know, it's not a code in China. It's more like people I give you a problem you are problem solver You solve the problem and often we give multiple offers to the engineers They say like I actually I don't know these questions. I don't know the spark and so on so forth However, I take maybe two hours or six hours to learn by myself over the weekend I get the answers right now here That is also a very important way. We're working. We are looking for the candidates like, you know in the real Day to day work or life. There's no way Like we spend so much time to train. Of course, we do the training the new joiners. Greenhands engineers to become mature step by step But what we are really looking for is like they really motivate to learn by themselves We really motivate to learn fast. They know how to ask questions, you know problem solving even We used to give a offer to our engineers who told us clearly, like they don't know, but he or she called our previous colleague who knows this. And then give him or her a lot of like suggestions otherwise and get it sold with a very limited time. Perfect. We hire them. That's what we need. Okay, so that's go back to your questions coding challenge. Yes. It's heavily impacted by generat generated AI chat, GBT, you know, your copy pass basically. You have a lot to answer, but we are not really afraid of that On one hand. We need the people who solve the problem quickly. That is one, like a fact. Like these candidates, if they know chat GPT or so and so, they solve the problem quickly. But on the other hand, we also have our way to their code to understand even they use chat GPT. But what is their really mindset of way of working? For example, I, let me give you one example. So we stand the coding challenge as notebook scripts with a real data set, a real models, real problems like our day to day work. It's going to face them, huh? And when the candidates send us the answers back, you can read their differences, such as those candidates tell you every single step of their way of thinking. Okay, I started looking at the question like this, and then I think like this, and then I get the answer. And then I maybe got even something extra. You don't even ask me, but I can do even better, give you more, like, insights from the data set of what can you do better. There's also candidates keep it really simple. You know, remove all the useless information, even the delete questions you ask. Just give you very to the point short answers. And then you see this kind of candidates, they are very to the point engineers. That candidates really like tell you all the journeys and give you all the logics to make sure like people can understand their communications and so forth. But there's also other engineers, you see the coding challenges, there might be three or four of them. all answers, all correct. But this answer is written by a language called Python. This is written by a SQL. This is written by a totally different language called Scala. Then you, you probably get a feeling like this answer probably is not like a one person's knowledge base. It's probably like piece by piece from there, here, there, here. put it here. You see? So actually If the coding challenge is really a good one to use real data, you really have the, you know, the logic behind that. You also have a full up session, you know, when you deliver this, it's not ending of the process. The process still have one hour Q& A. The first question I'm gonna ask a candidate, How do you feel about this coding challenge? Which task is the most difficult one, you think? How much time you take it? Is there anything you don't know but you learned by yourself during the process of limited time? And then what is the most important takeaways for this coding channel? You know, this is more like a comprehensive thing to help you actually still to see like what is this candidate's mindset? What is his or her way of working? You know, gave you this 20 percent of intuitions of feelings like if he or she is just an engineer gets a job done or he or she's actually a star with huge potential to keep us learning to be really have a really good logic to solve the problem and so on so forth.

TIM: Okay, so even though LLMs have come along and made it easier to solve the problems in that take home. There's still an enormous amount of value you get out of it when you include it with that followup process where you dig into it, because it's almost like the test itself is one thing, but it's like how they approach it. It's like the meta, the meta test or something where it reveals all these other things about their way of working.

QIANG: Absolutely, the process matters, it's not the final result. The process always matters. One thing I want to highlight here for the coding challenge is also, when we send in the coding challenge to invite the candidate to do it, we mention clearly three things. One thing is like this coding challenge has no time limits. means like you can take as much time as you want to complete that. thing is like this coding challenge has no skill set limits. That means like now there's plenty of tools available in the market to solve the same data problem. People often use Python. They can also use SQL. They can also use Scala or all the rest of tools, DBT, whatever. There's no limits. Do whatever you want. Whatever you're familiar with for that. You want okay last but not least. We also mentioned like there's three or four questions here They're all open questions based on the times you have you can choose just answer one question or two questions three or all of them you want. Yeah so this is also a highlight of the process matters because the process help you to In the end see the result and the future the best candidates for the feats for example often the candidate who sends a coding challenge in one weekend or one week, they pass the test. But if the candidate often take, if sometimes they take like a month or so on and so forth to deliver this, gives us a feeling like maybe this engineer is not like a deliver kind of person, as you mentioned, like make it happen. So, or like, are they really motivated for the job? Because if I see something, I really motivated, I can't wait for the next day contact to the people, to talk to them, to ask like the more information process and so forth. Right. So this is also very, very important part for like coding Chinese. Not just, it's just not a task. It's not like you gave me the right answer, but the process and the font answers, like this kind of things that the AI cannot help you, you know, AI cannot tell you, like, you need to get this done in two days. Right? Yeah, something

TIM: That's really interesting because yeah, one of the. Big benefits always of these take home projects has been that they're open ended. And so the candidates can kind of go down any route they want, but you've made it almost another level open ended in the sense that it's like, well, do it in a day, do it in a month, do it in any language, do one question, do four questions. It's all up to them. And so then it must give you a great variability in their approach, which just gives you all these extra data points. into, as you say, their way of thinking and their process. That's

QIANG: that.

TIM: interesting way of doing it.

QIANG: Exactly. And this is especially when we talking or communicating with our HR team or recruiter company help us to find the jobs. We mentioned clearly to them, like, please don't push the candidate to deliver. They have as much time as they want. Don't push them for that at all. So this is like something like when candidates open the code in China, they go, Oh, this is simple. But it's not simple. A lot of things to be careful to success in the process. And that's actually allies to the team culture. I think every leaders of the data teams, they have their own way of working. For example, my teams, often I'm looking forward to my candidates, all as entrepreneurs, often they come to tell me like, what's your problem, but they don't come to ask me. What's the solution? it's more like they come to tell me. Oh here What's the problem and what I propose the solutions like that and why I think in this way And my answer often is like give a try. What are you waiting for? It's okay to fail We fail fast, fail hard, fail strong and then after a bunch of fails we success once that's still faster Then like people tell you what to do. Yeah, so I think that's it's like an ecosystem of recruiting process like Culture really matters in the recruiting process so that we team everybody forming same way working even we are different color of people and then drive the excellence of the team to keep deliver the things because they cares they owns And they propose they deliver.

TIM: That's a really subtly valuable insight that you've given there, which is that The more connected the hiring process itself is to your work process or what you're looking for your values, your team, the better it is. And so often I see maybe more often in larger organizations where there's this weird disconnect, where it's like, I don't know, maybe the first two steps are handled by the talent team, and they're doing their own thing. And then there's a bunch of interviews. But the hiring manager, they're doing their thing. And it's just this weird disconnected thing where they're just doing the standard stuff, but you've really thought through it. And there's, yeah, all these little, almost like meta tests throughout the overall process. And yeah, the fact that you tell them not to push the candidate, because what kind of an entrepreneur would they be if they need to be pushed? They should be the ones doing it practically, isn't it? So that's really, really clever way of doing it, I think.

QIANG: Thank you.

TIM: One final question I have for you, Kyung, is if you had the proverbial magic wand, how would you fix or improve the hiring process?

QIANG: No surprise, you know, if I have really this magic wand, I probably will ask them to create a magic mural. And in the magic mural, no matter where you're using whatever process or AI or like what people are talking about, we want to reach to the AGI, right? general intelligence. Just help me to really save time to get what I need to solve the root problem. As I mentioned in the whole interview today, what I need is like, I really don't care where are they stands now as the engineer, cares about their mindset, their way of working and their potentials. you can tell me by the limited success stories they had in the past, even like people just graduated from school that never have any working experience and they, their CV might fool about like a school projects and so on and so forth. But you can tell me clearly like this guy is entrepreneur way working. This girl is fast, keep fast learning and also so fond of coaching her team's mates to grow together as fast as possible with her and those bunch of people they all have different colors, different personality, but that matter when we work together as one team, we are perfect team. That will be the things I would ask the AI, or ask the magic bone to help me to create a mirror so that don't need to manually go there to read their LinkedIn references find the clues, how they work, so on and so forth, except like their Twitter. Schools of diploma or companies that worked from Google, AWS, not at all. We hired so many good candidates. They are just graduated from a normal school, but they are rock stars or they never worked in Google or Amazon or Microsoft. They just want to work in the retail and they did such a good job. they motivated yourself, driving the entrepreneur, the team player, so forth. That is a big wish from my side for the Magic Wood.

TIM: I wish for the same thing. And I wonder if we won't have your wish granted in the next couple of years. I, I feel like we might, I think things are progressing so quickly that what seems like magic now will be reality in a couple of years time. So time will tell. Keon, we are out of time, speaking of time, and it's just left for me to thank you for joining us today, sharing all your thoughts, all your wisdom, all your experience with our audience today.

QIANG: It's with my pleasure, Tim. I also personally really appreciate for this opportunity, because I have very important takeaways from you, you know. So, and I never thought about this, but you mentioned, you pointed out, it's like actually, today we are talking only about recruiting process. It's one very important part of building a success team. I think I kind of like showed a proven track here. Like build a success team because my team was known nominated likes the top 10 best team. And I didn't point out to myself, actually, I build all of the things, including the recruiting process by ecosystem

TIM: Yes.

QIANG: My team has a culture the team culture is like one of the very important team culture is entrepreneur spirit. So therefore, actually, when we do the recruitment as the very beginning step, we already try to find the best candidate that really fits that culture. So that everything working together. Fine, right. So thank you so much for pointing this out. I always have the process but I never thought oh actually that flow Flows into my

TIM: Yes.

QIANG: Of the whole team is the ecosystem and the culture is the core of the team To make all the talents working together united here could keep deliver So therefore in the recruiting process, we already spent much efforts to find the people fits over there fits the way working Thank you team

TIM: Well, thank you for sharing that ecosystem and that very well thought out coherent process with us. I'm sure our audience learnt a lot as, as I did.