0:00 The muscle memory that I built for being an engineer has no use in the new world. I have to relearn everything in senior engineer role. My job is giving solution, my job is executing here. My job is like not giving the correct solution but making sure your team takes the correct 0:16 solution. The clarity is what differentiate. It's not the seniority, the ability for you or anybody inside your team to give a clear picture. This is what we are doing and this is a purpose why we are doing it. 0:28 Is it worth compromising on some of that foundation to go live faster with the product, to test the waters and to see if there's usage? As the manager, you should try to grow people at the cost of your growth. 0:41 Are you OK to do that? Joining me today is Sendhil Nalayapan, Engineering Manager over at Uber, and we talk about system design and designing systems for scale. 0:51 What are some of the ingredients that can make that a success? 0:55 Or how can you make it a recipe for disaster? 0:58 Lastly, we cover how to move from individual contributor to engineering manager and what are some of the trade-offs you'll have to make as well. So enjoy. 1:10 You've been involved in building systems that can scale to quite some users. I'm not going to name any numbers, I'm going to leave that to you. 1:17 But when we're talking about this recipe that is success, what are some of the ingredients that go into making this a success at scale? I think the biggest learning for me in Ingredients for the Success is trying to optimize 1:32 towards perfecting the road that you're putting forward. 1:35 That is basically try to architect in a way you build the layer, lay the foundations better. 1:40 When you're laying the foundations, take it a bit slow. 1:43 Understand what you're building, understand the criticalness, criticalness of the systems, the problems in the system and how we can deliver things across. Once you've laid that road, then you can execute at speed with a high quality so you can actually 1:57 go skyrocket. And once you're delivering, you have to make sure you're having enough signals that you're learning from the users. You understand what the user wants to use the system and you also understand what are the 2:12 trade-offs that you have made. And with these three things you collect the data and understand more and more about the trade-offs that you have made and what is the outcome of the trade off. The more you learn and more you 2:24 go into that feedback loop and shorter you make that feedback loop, it makes it much more easier for you to understand and then go faster. Is it worth compromising on some of that foundation to go live faster with the product, to test the waters and to see if there's usage? 2:40 That depends on the problem that you're solving, right? 2:43 For example, in one of the bigger initiative that I'm taking right now, I am first I started with the POCI, started ask give the team like, hey, this is clearly what we are trying to achieve. But let's try to solve this 2:55 problem in a much smaller scope. Try to find whether the solution that we are putting out solves the actual problem. 3:02 And once we try to put that forward, we figured out like, OK, this there is a problem and we are putting a small solution to fix it in a small scope area. And that small scope area is actually fixing the problem. And we got the results from it. 3:15 Now this, I can make it faster, I can take shortcuts, I can make sure like, OK, the impact is going to be much more minimal. 3:22 So I can try to put something outside, learn from it, get in this data. In addition to the data, you also get the intuition. You have built a lot of systems, you understand how things should work. 3:32 You compare these two things and take a decision based on that. 3:37 Once you've done that, then what you're going to do is like scale the system to the next level. OK, now you're scaling from few 1000 users to a few million users. 3:47 When that scale happens, now you have to think very hard about your budget, your SL OS, your SL AS, and use that mechanism and the feedback and the feedback that you got from your first launch. Take that data, combine them together and construct an architecture such that the next 4:04 two years that you're going to build, you have to go faster. 4:08 At the same time, you have to ensure what you're delivering makes meaningful impact. So compartmentalize these things, split into smaller chunks of problem and then try to go through that. That helps a lot. 4:21 I can see that I am a fan of this way of thinking. 4:24 Starting small, going live with a good foundation and good is kind of subjective throughout the organ throughout experience. 4:32 And I've had this conversation similarly on the podcast as well. Then a lot of my audience were like in some organizations, you cannot afford to take this approach to go live and to look at your architecture and then re architect for the next magnitude of scale. 4:45 But I'm not sure why I'm curious how you view that. 4:48 Can you, should you architect for a certain level of scale already quite far ahead in the future or I don't know necessarily if that's the right approach? 4:55 I think the trade-offs matters, right? 4:59 There are different kinds of system. 5:00 Even the way is like right now I'm architecting one of the main pathways Uber uses. So when I'm doing that changes, I cannot go and change something radically. 5:12 And if that causes an incident, it's going to be a huge impact on the business. Yeah. 5:16 So I can understand there are like some legacy parts of the system that I have to change. Now in that scenario, the way I look at it is like I try to considering a walled garden. 5:27 I have a feature, put a wall around it, put two sentries on outside, and these people see the evens that comes, goes in and comes out without evens. I can map the features, I can map the flows, I can map the funnels so I understand how the 5:42 data goes through what is coming out of it. 5:45 Now I'd be more confident to do some changes inside my garden. 5:49 I can start from a small square and change that particular thing and then I can go step by step. When I am building that garden, I cannot when I am revamping that garden, I cannot put the flooring as well as plant the flowering at the same time, 6:01 right? So I have to take my parts and try to build each and every step. 6:06 And with that I can understand like what I can change and because I made it very clear that this is what I want to change and I also made it very precise. 6:16 This is a small area that I am going to take as a first step. 6:19 The clarity gives a lot of confidence and the smaller feedback loops or the shorter feedback loops that I'm creating also gives a lot of evidence for me using these two, I can comfortably do the migrations. I can comfortably go step by 6:32 step and then evolve the product go chats where I would optimize. 6:37 Yeah. Have you seen people that are less experienced struggle with this? 6:42 Because if I have a big change in a big feature, especially if legacies in the mix, and especially if there's not really a lot of comfort and confidence sometimes about releasing, then making it smaller is the way to make it digestible. 6:54 But making it smaller is also quite a skill. 6:57 From my perspective, the biggest change that I'm seeing is the clarity. This what differentiate. 7:02 It's not the seniority, the ability for you or anybody inside your team to give a clear picture. 7:09 This is what we are doing and this is a purpose why we are doing it and what and why there's a lot here. 7:15 So once you give that clarity, that makes it easier, irrespective of your experience, when you know clearly like what you're trying to solve and how you're trying to solve, it makes it easier for everybody, right? That's that's how I work. 7:28 Yeah, I've had the most experience in existing code bases and then trying to figure out how to add something on top of what we already have. That clarity is quite important. 7:39 And I noticed myself in some experiences where I start from scratch. I struggle, right, because I start with a blank canvas and I know if I have something existing where things make sense based on certain conventions and certain decisions have already been made. 7:53 So especially in the use case of starting from scratch, whether you already have a big organization or you're starting literally from zero, I feel like a lot of decision making there is quite interesting, especially when we're talking about then 8:04 designing for scale. I feel like people are quite opinionated and getting those people on board with regards to indeed having a very good purpose, a goal and clarity on where we want to go, I feel like that's definitely part of the ingredients for success there. Yeah. 8:20 I think as I said before, like people have a muscle memory of what they have done before and what worked for them. 8:27 And when you actually start with a blank canvas, they try to solve the solution with the same lens that have they have been wearing. And that is as a manager, my major role I feel is like you have to remove that lens. 8:39 You have to make them see the entire picture. 8:42 They are starting at a blank canvas and this canvas has a different story, not the same story that they have solved before. And optimizing that, understanding the strengths and the growth points for a person helps a lot. And that's how I see the systems 8:57 in architecting the systems. It's like, OK, this is what we are trying to do. This is what This is why let's go and solve that, but let's keep our assumptions outside of it. You have solved for 10 million users before, but right now my solution is going to focus on 9:13 1000 users. Let's try to optimize towards there. You can get the learnings that you have from the previous thing and put it as a part of the input in that equation. But when you are actually solving the equation, let's try to solve for the particular use 9:25 case and then move forward. Do you see then a difference between the people that are more senior, that have more experience under their belt where those assumptions kind of steers their decision making, sometimes even in a wrong 9:36 direction versus the more early in career people that actually sometimes have a blank canvas? I think it it does in both the directions I've seen, especially people who joined the company newly, they are very enthusiastic. 9:50 They try to put all the great principles to work. 9:54 But in an ecosystem where there are some constraints, you cannot put every good principle to work. 9:59 And the experience matters a lot because experienced people comes with context and some people can easily adapt and change. 10:08 For some people, it's going to be difficult to adapt and change. I think the ability for you to adapt matters a lot and the ability for you to think, given the constraints, given the limitations, how can I change things that differentiates a great team. 10:23 This is a good team. I like that a lot, Yeah. 10:25 When we're talking about speed of development and rate of going live, going to production, I feel like there is this trend and I noticed it first in games where games used to be on disk, you would get a complete game. All the bugs in the game are in 10:39 the game because it's on the disk. 10:40 So you can't really do anything about that. 10:42 But when we went to kind of a software first distribution of games, you can get a day one release with a lot of bugs and then a day 2 patch that solves a lot of those bugs. 10:51 And I feel like that approach sometimes is similar when going live with an MVP to kind of test the waters and to grow from there to get data usage data. I think it's a smart strategy, but I'm curious from your perspective, in the foundation 11:04 that you build, what is some of the stuff that you can compromise on and what should you not compromise on? 11:10 You mean the architecture that I'm going to build? 11:12 Yeah, Yeah. I think based on the what I can compromise on some parts of, for example, if the system has to have this much amount of latency and I'm trying to write a POC for now, I can compromise on that metrics. 11:28 I can compromise on some parts of like single door decisions that I'm making the decision that will be reversed, that can be reversed that without a large amount of effort. 11:40 I'm OK with taking that decision and moving forward. 11:43 So we can launch the production, get user feedback and then try to do it. Especially when we're trying to solve something Greenfield, it gives a lot of, we need a lot of data. And while even you go to production, you can understand. So that's how I optimize myself. 11:59 And there I would push more and more on the velocity side of things and understanding we are making some assumptions, tracking those assumptions and writing them down at some places. So in the future something happens, we can always go back to the assumption and correct 12:16 the assumption. So my focus in that area is like having the shorter feedback loops, understand the system, understand the user, and then go and iterate on that. 12:26 That's how I would like to operate. 12:27 Yeah. When you've then proven that, let's say you're building a proof of concept, one of the biggest trends that I've seen is that people then take the proof of concept and say this is our foundation now we build on top of that. And then because you've made 12:39 those trade-offs early on, it can be quite an unstable foundation when we're talking about actually going the distance, which is a challenge. Some organizations and some engineers get it proof of concept to test the waters. 12:52 Then OK, if we really believe in this, we have to invest in a better foundation. But for those that don't, it can be kind of shady down the line with the rest of the complexity that you build up. How have you seen people battle complexity, specifically when it comes to scale? 13:08 I think once you go to the production, the reason why I made sure you have to write all the assumptions together and then clarify what the hypothesis was and what is outcome was. 13:19 Then you can feed that as a data into the system when the complexity grows. Now you have to take this and your current architecture that you have and then sit together with the entire team and understand like right now we are scaling this. When you are scaling what you 13:36 are optimizing for, are you optimizing for latency or are you optimizing for more users and more impact that you want to generate? You're OK with users waiting for a second because you're not OK for users waiting for a second. 13:50 So put that complexity in and then specifically solve that complexity and extend your solution for that complexity. 13:57 Now you have a system that works well, that can give you shorter feedback loops, that can give you data that you want and you also added in the extra constraint. 14:07 Now you launched the protection and that's where you will understand what is happening inside the system and you will get more data out of that system and you continuously have to revalidate those things once you try to have those loops. 14:20 And a well functioning team or a high performance team always looks into this loop, understands where the problem is, and then tries to fix the problem. 14:28 And once you make it as a culture of the team, it becomes easier for you as a manager to look on the next target or to solve for the next complexity. So that's how I would optimise. 14:38 And that's fair. Like I've seen things often break in that direction where like you have ABOC and then just launch the BOC outside and say like, OK, I'm going to launch this. I'm going to apply the same principles. You cannot do that always. 14:52 And you have to find out like what is differentiating you and what is the complexity that you're solving for. 14:57 And that's where I, I always fall back to the clarity. 15:00 If you cannot give the clarity of what you're solving, it makes it more complicated. Yeah. 15:05 This writing down of assumptions and kind of validating those assumptions and then keeping those in mind with regards to the trade-offs that you make US engineering manager, from whom do you expect this? Is that a certain level of seniority? Is that engineers? 15:18 Is that product? Is that the team as a whole? 15:20 Does it matter who does it? I think one thing I like in all the organizations that I've worked with is the ownership. 15:26 Yeah, I would like everybody to take ownership of what they're putting into the into production. 15:33 So I would not call it out as like this person's responsibility or managers responsibility. 15:38 This is everyone's collective responsibility. 15:40 If I'm not doing my job correctly, it's your job to point it out. So that's how I would like to set a team. So everybody should have the responsibility. The moment you make a, this is the responsibility of staff engineer, then it makes it like 15:54 OK, tomorrow some senior engineer will come and blame staff engineer did not write. I don't want that. 15:59 If a staff engineer is not writing, then a senior engineer should push. If nobody is writing, then the manager should push for writing that or he has to take the ownership of doing that. So the clarity here is as a system, if you have to work perfectly, we need to have an 16:14 operating manual and that operating manual should have all the assumptions. So try to build that assumptions and try to run write it down. And anybody who's pushing the code into production is responsible for writing them down and validating that. Yeah, I like that a lot. 16:30 I'm I'm very big on talking to people about their behaviour, right. I don't like a big governance framework. A lot of the policies in the way we do things, I do think they're necessary and healthy. 16:41 But when we're talking about things like this, this is a perfect example of we can address each other's behaviour right and have a mindset and shared ownership as a team. 16:50 The problem is if no one does it, because then actually no one does it. But if people are kind of keeping this top of mind, then you can address people's behaviour and your behaviour is not necessarily who you are. 16:59 So you can detach those two. And sometimes we also act in a way that we think we do right, but it might not be the right way. Yeah, as we execute, the thing that I'm also seeing more is like my role is not to talk about day-to-day stuff, but try to build a framework and give 17:18 the framework to the team so that execute that framework. 17:21 It's not a close ended framework, but it's an open-ended framework. Like hey, I'm running this project. What are the things I should look into? What are the areas that I have to do? How should I operate on this? 17:32 If the project is in this state, what should I be knowing about it? And this is an open document and living document inside the team. So anybody can come over and update what they think is optimal for the team. 17:43 So we have this bunch of checklist and a bunch of lists is learnt for every stage of this project. 17:49 And using that as a manual, anybody inside the organization can take that and run a project. It gives more clarity and it also gives a mental piece like, oh, am I missing something? 18:01 Is there something I'm missing? So giving that clarity actually helps for us to run multiple projects in parallel and build a very high performing team. And that's where like my mind shift also happened, where like instead of tackling every design 18:16 decision, why cannot we make a framework out of it? 18:19 And why cannot we help the team so that they follow the framework? For a first year and a half, I was like, oh, we have to do this here. 18:27 We have to do that here. Have you thought about this? 18:30 Have you thought about this Later it changed into like, OK, have you followed the structure there? 18:35 Do you think are we missing anything? 18:37 And at the end of every project, we used to have a retro where we go through this project list and the items that we have to do and then change those things across accordingly. 18:46 Whatever we learnt from it, it's a very fruitful process for us and it helps a lot. Yeah. 18:51 Do you get fulfilment out of seeing high performance teams or also out of kind of more being in that team because you've had experience of both? Yeah, I think personally I would like both of course, every every team and every team that I'm 19:09 building from the ground up. I tend to see how the culture is changing. And lately I've been like super focused on building high performance team and building a team that kind of like not only delivers results, but questions back and comfortable in questioning each other. 19:26 But they all disagree and commit. 19:29 That's the culture I want to sit inside the team and that's a culture I'm more and more enjoy working as from my early time itself. So that's where I'm optimizing myself. Yeah, you see that. 19:40 I want to go deep into one of the examples you gave, mainly because I don't think I've had many conversations specifically about latency at a certain scale. 19:48 So when you say OK, we make a trade off and we say the user can wait for X amount of seconds, I don't know when I would play that card or when I would have that opinion because I haven't built products at a certain scale. 20:00 At a certain scale of a magnitude of users, your latency will really make a difference somewhere down the line. 20:06 But even me saying that, it's still an assumption. 20:09 How do you navigate those conversations when it comes to users actually experiencing the app with regards to latency and then also the efforts that it takes to optimize for that? 20:20 I think latency asset constructors, we are giving over importance in the industry right now. 20:27 There are some aspects there. If you want to ensure, for example, a compliance related stuff, if you want to ensure like your code is very compliant or things are very compliant, the things that you're showing to the user at that point of 20:42 time, instead of showing a wrong information to the user, you have to show correct information that takes some time, it's fine. 20:50 So it's basically talking about strong event, strongly even, sorry, eventually consistent, which is strongly consistent, right? And the way I think about it is like if the data has to be perfect and it increases the trust for the user. I work in the payments to mine 21:06 in Uber which means like I cannot show wrong data and I also have to show correct data at the correct time. 21:12 Yeah, engineering at the highest. 21:14 Level. So how can you optimize? 21:16 Because if I start showing wrong data it's going to break the trust for the user. So what I'm trying to optimize here is like if my latency is higher then try to not show the data. As long as you have the correct data, correct value is rare. While in the experience side of 21:32 things, how can I prefetch or preload this data? 21:35 How can I ensure that it is available when the user goes to that particular part of the screen? 21:40 That's where I would optimize and there are a lot of places where you can prefetch the data, where you can make sure the data is available at the correct amount of time. 21:49 So you can use this philosophies and try to optimize yourself based on the existing use cases. I come from a web background and the preloading and prefetching are like greatest way in which you can speed up JavaScript sites right in your websites. 22:04 And I was like when I was an engineer, I was like super mad about micro benchmarking and the micro optimizing things. 22:10 So the way you can optimize is chunk your data, make sure you prefetch it and preload it as you need it. 22:18 So that's how I would optimize there. 22:19 So latency in this contract for me is like I'm getting the correct data set based on the current context and then I'll optimize it by showing it differently or loading it differently for the users. Yeah, I feel like this also 22:33 reflects again the clarity that you have in your product, right? 22:35 If you are in the payments domain, you don't, you don't have much room for error, right? It's it's quite black and white. 22:41 Either it's correct or it's not correct and not correct has a lot of downsides. So we need to go for correct in that case, which means that optimizing for latency comes second when it when you're comparing to showing the correct data. I really like that part. 22:54 We also talked about ingredients for success when we're talking about building systems at scale specifically, what are some of the ingredients that you're saying? 23:03 Man, this is just a recipe for disaster. 23:05 This is really what you should avoid or some of the pitfalls that people fall into. It. 23:10 Depends. Yeah, of course. 23:12 Like, well, coming to this interview, this podcast, I was like thinking, I should not say it depends. 23:17 You can say it's very difficult. I'm going to ask all What? 23:19 Yeah. I think the biggest recipe is like the opposite of what I said, that clarity. 23:26 If the team is solving something without knowing why and what they are solving, what is usually granted. 23:31 Like you will know what you're solving, otherwise you will not go for solving right? But why is very important. 23:36 And if the why, the clarity for the why is not there, it makes it much more complicated. Then people will not see the mistakes that as a manager I'm doing or as a staff engineer somebody's doing. They will not see that mistake. 23:48 They will blindly follow these things. 23:50 And I think it makes your engineering process itself more dogmatic when you don't understand why. 23:55 So giving that why reasoning and making sure you understand the why correctly gives a a lot of it over the other things. 24:05 I think that's a major part of it. 24:07 The the next thing that I would say, which kind of like makes it more disastrous, like people having very strong opinions on some patterns and frameworks. Oh, I'm going to just use this particular pattern. No matter what the solution is, 24:22 I want no matter what we are solving, I want to use this particular pattern. I remember like in here in, in my previous office, there used to be a single code base for calendars. And this was a web calendar thing. But the entire code was written in Scala, OK, pure functional way. 24:42 Just because, just because I, I vividly remember like when I joined the company, nobody wants to touch that code. 24:50 Yeah, everybody was saying like, no, I don't want to touch the code. But the code was working perfectly. But I we cannot do any changes on top of it. Out of fear for like breaking it or? Nobody understands. 25:01 Yeah, it's written in a pure monadic Scala way of very functional way, and it makes it much more complicated. 25:08 It's probably very elegant, yes. It was elegant, but it took like 2 1/2 years for us to remove that NTA code. 25:17 That code was in existence for 2 1/2 years and we removed it and replaced with some library. Then people were more comfortable in doing changes on this thing that we're doing. 25:28 So I think that opinion and trying to build things in the way that I know, rather than the way that it fits the solution or fits the expectation that we have breaks a lot of things and that's where disaster happens. So I tried to avoid these things. I tried to figure out like how 25:43 can I optimize this and identify this information beforehand. 25:47 As a manager, I would have kind of how to say it, a vision or additional information that I can. 25:57 I'm talking with this people every week or every once in two weeks. And I understand what they are solving and what are the things that they are missing in understanding. And that gives me an edge and it also gives me an opportunity for me to explain it more vividly 26:12 and clearly to the team. Hey, this is what we are trying to solve. I understand that this has worked, but let's not use the same stuff. 26:19 Let's try to think about it again as you grow, you are enough understand like it's not your job to go and pick on every design decision and change those design decisions in an optimal way. You have to let people fail and 26:35 based on that failure, we will learn from it and that's how we have to optimize. I, I want to zoom in on those failures, but first, when you talk about engineers specifically within Teams, I'm wondering what your advice is on 26:48 if they notice this is kind of the recipe or the ingredient that this team has. One person that says I want to do Scala, that's what I'm good at. 26:56 I think this application needs it. 26:58 Or in any case, I think we can go faster because of this. 27:01 That has really a strong push towards something maybe not fit for purpose, but definitely based on experience, a clear preference. How should people within that team kind of navigate that when it comes to finding a solution 27:13 that is more fit for purpose? I think it's asking the right questions. It's figuring out where the problem, how you're solving the problem and understanding like what is the long term life of the problem? 27:27 Are we, are we solving, solving just for a short term, just for this POC, or are we going to solve it for a much long term? 27:34 Understand that nuance and applying the right tool for the job always matters based on the context that you're working in. 27:43 That's how it has to work at some point. 27:46 You also see like you have a team of great go engineers and India team and India organization is using just go in that scenario. Yeah, that's that's that's that's a point where you cannot go and have a conversation. 27:57 Hey, go and have use rest trust is very good for this particular thing. It's a right goal for that job. 28:01 Then you try to work on what you have and try to optimize yourself towards there. But if your solution actually needs safety and concurrent system rather than just a concurrent system that is highly performing, then you should fall back to that. You should advise your team to 28:18 go in that direction and team should advise you to go in that direction. So that's how it has to work. 28:23 Exactly. Then you look at the what the language brings in the rest versus go example and if it's fit for purpose for what we are building specifically. 28:31 I like that a lot. When I made the switch from individual contributor software engineer hands on to being responsible for product, I had a lot of conversations with engineers with regards to sometimes a difference in opinion. And then I thought to myself, 28:46 should I prevent this person from doing this or should I just let it go? They can learn from it. 28:51 Maybe they fail, maybe they succeed. 28:52 I think they're going to learn from this and it's not going to work. But should I prevent or should I let them make their own mistakes? 28:58 And I feel like you from an engineering manager perspective, you probably have similar conversations. 29:03 How do you navigate that? Yeah, I think this is one of the question that comes very top in my mind. 29:08 Like whenever I'm having a conversation where I see a person just very highly opinionated and wants to drive in a certain way, they are very comfortable in that they know what to expect and they can do wonders if they go in that direction. For me, it all goes back to the 29:22 design phase. There you put you lay the guardrails, you put the road and you ask your team to travel on it. If you're building a car, you should put a road. If you're building a spacecraft, you should do what where it goes and what it works for it. 29:38 And that's how I would optimize there where like my major focus is like these are the constraints that we are going to build. The constraints is going to be a living thing. It can change at any point of time, but let's assume these are the constraints and let's go in 29:51 that way. It kind of increases the speed and it also make sure the quality is higher because you're going to work within the constraint. 29:59 People know what to do and which makes it much more clearer. 30:03 The clarity gives speed. So when you go in that path, when you go in that trout, it makes it much more easier at point where you have to think before taking a decision. 30:13 These are like irreversible decisions that you're making. 30:17 It's very similar to Russ versus code. 30:19 It is irreversible decision. You go one way and you cannot come back the other way unless you rewrite the entire code, right. So in that world, you have to stop and you have to think about like, what is happening. 30:29 You will have speed Breakers at some time and you have to go for, you have to go for refilling your gas for your car. 30:36 So you have to park it there, think about what you're doing, and then come back to the road. So that's how I think things will change a lot. And my role there is like strategically pushing and understanding what is happening there. How can I help you there? 30:51 Yeah. At some places you have time and leverage to let them make the mistake and then learn from it. 30:59 At some places you might not have. 31:01 And that depends on the current context that you're working on. 31:04 And based on that, we'll make the decision. 31:05 Yeah, I like that a lot. I feel like that's the the muscle you will develop in having more of those conversations where you'll be able to gauge in this high risk situation, can we afford to take this risk, right? 31:16 If we disagree or do we actually need something else or a third opinion or I we, I do need to act basically. 31:23 And in other places, it's interesting because I feel like in the managers that I've had, the people that will micromanage, they are trying to prevent their people from making those mistakes. And I felt that. 31:34 And from my side, as soon as I noticed myself doing that, I'm like, I have to step back and be like, this is the behaviour that I don't want to reflect. But it's quite interesting to kind of manage that and make those trade-offs. 31:45 I learned a lot from that. It's a lot of fun, to be honest. 31:48 Yeah. You grew from IC where you have build systems at scale and I can see you have a vision on that as well. But now you're in a role that is I think less hands on. Just to the fact of you being engineering manager, what are some of the trade-offs that you 32:02 considered while going down that career path specifically? 32:06 When I first become an engineering manager, I told my manager at that point of time, look I cannot stop coding. 32:13 Yeah, this is me, Yeah. This is me. 32:15 I cannot stop coding. I'll continue to write code no matter what and as my number of reports keep on increasing, the number of dips that I'm going to create fell down drastic. 32:27 Then I realised this is a very different job. 32:29 I thought it's just an extension of being a senior engineer. 32:32 The muscle memory that I built for being an engineer has no use in the new world. I have to relearn everything in the world of in senior engineer role. 32:44 My job is giving solution. My job is executing here. 32:47 My job is like not giving the correct solution, but making sure your team takes the correct solution. 32:53 You don't have an authority. You have an authority, but you don't have. I don't want to use that authority often. That's different. 33:01 Yeah. And so I would love to understand what is happening and how can I change things. 33:09 How can I tell them they in a way that is not so direct when I'm an engineer I could say like this is bullshit. 33:15 It's. Wrong. 33:16 You should not go in there. Let's go this way. 33:18 Yeah, you can take ownership. Yes. 33:20 But now I cannot do that. I have to be more and more practical, right. Very delicate. 33:24 Yeah. And so that is the major change for me. That muscle memory that I built has completely. It's completely useless. 33:32 Yeah. And you were confident in that one, the muscle memory, right? It's a skill. 33:35 You take pride in it now, my. Role is like, OK, this is what is happening. I would have solved it in this direction, but I cannot say this is the direction because people in my team are great. They know better ways than me. 33:51 So if there is a mistake, my instructor fall back is like let's not call that mistake, but let's ask questions. 33:59 So I go for probing. I ask them more questions like hey, what is happening here? Why do you want to go in this direction? The more and more I ask question, I understand major major of the time if because my team is super awesome majority of the time I can get back to my 34:19 place and I think send a leadership you should be thinking better. They have a good solution. 34:25 But sometimes my engineer sense also like takes a pride or I found that mistake out. I can still do this. 34:35 But it it depends on the context. 34:37 And that's where I said like the my major fall back scenario here is like always question myself before I let out the Commission that I have or let out a solution. 34:48 It's not very natural for me. I'm still struggling. 34:50 I'm still learning from it. Every meeting that I go and I tell myself, suddenly you should not tell a solution. 34:57 You should help them do it's a solution. 34:59 You should not tell a solution. That helps me a lot and that's where I would allay to. That's the biggest and most difficult change for me and it's it's happening. 35:08 I think when we go back 20 years ago, I wasn't in this field, but this is all hearsay. What they've told me, the natural career progression for a very senior engineer, the only way for them to reach a new salary scale. 35:22 Typically when we were talking about compensation, is to go manager. That was the only path. 35:26 Nowadays, we have a different route. 35:28 You can stay icy and you can go very deep. 35:31 And when there's big organizational risk with regards to what we need to deliver, it goes to you, right? 35:35 When the stakes are high, goes to the most senior people that we trust because they've built up the credibility and they know their stuff really damn well, which means you have options now. Now there's a train of thought that says the very good IC people, they should stay IC, 35:50 right? That's their core strength. 35:51 You already mentioned that going into engineering management, there are a lot of skills that you need to learn and what you have confidence in, it's not going to be your main skill anymore. But my opinion is that there is still tremendous value for engineers to still go in that 36:04 career paths. And me as an IC, as an engineer, I see myself as if I have to label myself. 36:10 I see myself as an engineer. I would much rather have engineering managers or people in leadership positions that know engineering, that have gone through this, that know their stuff basically versus the experience of having people that 36:22 don't know that some conversations are just different or I have to emphasize certain nuances that people don't get automatically. It puts more stress and emphasis on me. But also those are the leaders of the company. So I do have to put my faith in 36:34 them, which means now for people to actually enjoy what they do, they might stay IC. But I think when we're talking about the good for an organization, I think some ICS also do need to move up to managerial and leadership positions. And I'm wondering how you see 36:48 that when people are actually making those trade-offs in their career? I think 2 paths to this thing, 1 is like the ability for your team to trust you drastically changes whether you have been a man, whether you have been an engineer before or not. Yeah, like magic. 37:05 Yeah, I've seen this pattern a lot where like non engineering person goes and beats a group of engineers, it becomes more and more difficult for them to establish norms, certain conditions or understand what they are solving. 37:20 And when the understanding goes away, the trust erodes. 37:23 On the other side of the spectrum where I would I would like more and more engineers to go to be managers and the one core skill that I would ask them or one core question that I would ask them or two things I would ask them is 1. 37:40 Whether they are a BFS engineer or a DFS engineer. 37:44 If they are BFS engineer, then I think they suit very well for being manager. If they are DFS, they have to change their approach. What does that mean? 37:52 BFS. DFS. 37:52 Breadth first search or death first search? 37:54 I gotcha. If I I want a manager should be Jack of all trades, kind of a position where like he has to understand multiple things and have more context on multiple things. So he has to go breadth wise. 38:06 But as AIC you can go depth wise, you can go deep down into one single thing and then you can be excellentizing that scale, right. And you can scale from there too. If you cannot get overview of the system and you are so strong in one of the domain, but not so 38:22 strong or doesn't have more context in other domains, then if you want to become a manager, I would advise them to go breakfast. I understand all the other areas. If you're a back end engineer, know how the mobile works, know how the web works, right? 38:36 So that is the first thing. And the second thing is aspiration for you to grow other people as you become manager. 38:43 It's a premise structure, right? The growth in the manager's side, it's pretty much limited. At some places it is like, hey, this person has to leave so that this person can go up, right? 38:55 It's just how it is, limited spots. 38:57 But as an IC, your growth is like, hey, I have to deliver this X&Y, then I can grow. That's how the impact framework and everything is defined. So given that given that thing, as a manager, you should try to grow people at the cost of your growth. Are you OK to do that? 39:16 I've seen a lot of ambitious people who likes to grow faster, who likes to grow in a shorter period of time. 39:23 If they become a manager, are they willing to take understand Like, OK, my growth will be standard to an extent. 39:30 I'm not saying I'm not making manager as a gloom and doom thing, but my growth is standard for some time in that. 39:36 Meanwhile, I'm happy in promoting other peoples in the organization. If you feel that happiness, if you make them grow, then that's a hint for you. 39:45 You can be a great manager. The reason I become a manager is because there was one person in my organization who did extremely well. He's one of the most awesome engineer that I've seen and he has some problems with communication. I helped him in the 40:04 communication. I sit with him, I had conversations, I drafted something and helped him towards communication and he ended up being a senior engineer. 40:13 And when he become a senior engineer, I was so happy that I got senior engineering promotion. 40:18 So that is the thing that kind of like made me OK, I'm doing in the correct path, I'm doing what I'm happy about. 40:25 So let's go and do that. So making others grow and making sure you understand the breadth of the organization that gives more and more leverage for you as a manager. 40:34 Yeah, I like that a lot here in the role that I have. 40:37 We don't really have much hierarchy, but there was an opportunity for me to become teamly, to be responsible on a very small scale for people's development and for their goals and to challenge them and to make sure they're happy at the 40:49 org, be responsible for that. And I did that because I thought, OK, this is my like micro skill path to an engineering manager. And then I learned from that and I like to be a busy bee. I have the podcast, I have client assignments that the Steamly thing works council. 41:04 I noticed myself automatically, even though I thought I wanted this de prioritizing some of those conversations or some of those developments and prioritizing my own growth where I could have the most growth. And the way you mentioned it 41:16 right now, it really resonated with me because in the end, now I'm no longer A-Team lead and it was consciously due to my own behaviour. I thought I wanted this, but apparently I want this other thing even more. 41:26 So I can want both things, but I have to decide because time is finite. So I feel like right now is not the right time for me to have managerial responsibilities to go into an engineering manager career path. 41:37 Because apparently when it comes down to it, I prioritize my own growth, at least right now. It doesn't mean I didn't enjoy having those conversations. It doesn't mean I didn't enjoy challenging people and then having them see results based on 41:50 conversations we've had. It means right now is not the right time for me. And I had to realize that. 41:56 Yeah, that's a good decision. Yeah. 41:57 But do you have aspirations to become a team lead again in the future? I so I see my career path as quite open-ended because I like the IC path. 42:05 I'm now responsible for developer experience and kind of an IC for an IT team, responsible for product within that team specifically. I like a lot of things, but the way you put it, with regards to breath, I do have that and I do enjoy having conversations with people. 42:20 I think I have developed, especially throughout this podcast, a lot of good communication skills. 42:26 But also in talking to a lot of people, I can give different angles and perspectives for people, either to challenge them or to help them. And I do enjoy seeing people achieve results And kind of even though you're not directly 42:38 responsible, you still have a team and you feel you take pride in that team. And if they do well, it just feels good on a different level, yeah. 42:45 Yeah, that, that gives you a very good satisfaction. 42:48 Yeah. And it's kind of, it is kind of like fulfilling experience to see things go and things end. 42:57 Yeah. And you have the only thing that you have done is just help them or guide them, just show them some direction, all the work they have done. 43:06 But somehow you feel like, OK, I've done a good job. 43:09 You did a good job. On my back. 43:10 That's it. Because no one else is going to do it. Yeah, that's one thing, right? 43:15 Like the moment you become managers, the amount of appreciations, appreciations or celebrations that you're going to get is going to be much more minimal. 43:22 It's going to drastically reduce the small thing that you're doing. Have very long time for you to get the benefit out of it. A small change like removing the 30 minutes meeting or something like that, it's good for people to realize the benefit. It's going to take a longer 43:40 time. As an engineer, you can immediately see the change in production. 43:45 And the other thing is like the moment you take some decisions, you would always have push backs, a lot of push backs and a very limited amount of acceptance, right? 43:58 So you need to be ready. You have to have a thick skin for that. You need to be ready. 44:02 Like, yes, whatever you're going to do, it's not immediate outcome. The appreciations is going to reduce at the same amount of time. 44:11 You are expected to appreciate other people a lot. 44:15 So that's kind of smaller nuances and smaller trade-offs that you have to take when you retrospect yourself. 44:20 That's scenario like you where you also have to be OK with it. 44:24 Yeah, yeah. Yeah, definitely. 44:26 I have really enjoyed this conversation, Sendhil. 44:28 This was a blast. Thank you so so much for coming on and sharing your perspective. Yeah. 44:33 Thank you so much. Thanks for the time. 44:34 Yeah. We're going to round it off here. If you're still listening, let me know in the comments section what you thought of this episode and we'll see you in the next one.
Community discussion