Episode Summary
In this episode of Emerging Founders with Manav, host Manav sits down with Shivani Mooli Swaran, co-founder and CEO of Ember Robotics. Shivani shares her journey from her early days studying electrical engineering, to her pivotal roles at startups and as a software engineer at Tesla Autopilot, where she specialized in optimizing camera performance for autonomous systems. She discusses the progression of autonomous driving technologies, the real challenges of scaling camera systems, and why processing data in real time is essential for safety-critical applications.
Shivani explains what motivated her to leave a secure role at Tesla to launch Ember Robotics, identifying a persistent problem in robotics sensor diagnostics that she believes is finally solvable thanks to recent advances and decreasing costs. She paints a picture of the robotics landscape—debating the near-term utility of single-purpose robots over humanoid robots, and underlines how the advances in adjacent tech, such as AI chips and real-time edge processing, are benefitting the whole sector.
The conversation also dives into Shivani’s experiences going through the prestigious Y Combinator accelerator, the fundraising journey, and the unique value of founder communities. Beyond her professional life, she gives a personal glimpse into her hobbies and the personal philosophies that drive her as an entrepreneur.
Transcript
Manav [00:00:00]:
Looking good. Boom. Hello, everybody. Welcome to Emerging Founders with Manav. Today's guest is Shivani Mooli Swaran. She is the co founder and CEO of Ember Robotics. Before starting Ember, she worked at Tesla Autopilot for about five years where she focused on optimizing camera performance for autonomous systems. With over more than five years of experience working for camera systems for autonomous robots, Shivani co founder, Ember. She found a rare gap in the market and she decided to pursue this and create a robotics company that will solve this problem. I'm really excited to have Shivani on the show. Shivani, how are you doing today?
Shivani Mooli Swaran [00:00:50]:
Doing really good. Excited to be here as well.
Manav [00:00:53]:
Awesome. So before we talk about Ember Robotics, I want to know everything about you as a person and I want to know where. How did you get started in your career? Like, what was your first job? Or how did you get into robotics? Let's start there.
Shivani Mooli Swaran [00:01:10]:
Yeah. So I actually came in from electrical engineering. That's what I went to college for. And I think pretty quickly, freshman year, after burning a bunch of circuits in the lab, I discovered that maybe fabbing circuits wasn't for me. But what's really cool about electrical engineering specifically is that digital signal processing is a huge kind of research category within that, and that has a lot of applications to image processing and computer vision robotics. So I actually started working on some drone projects with IEEE while I was in college. That led to my first internship with Caterpillar, doing some drone R and D work for tractor motion detection. And then that kind of spiraled into more internships where I worked at NASA the summer after, and then at a startup called Simbi Robotics, where I eventually went full time.
Manav [00:01:53]:
Got it. So your experience, I'm guessing, in electrical engineering, it's more like dealing with hardware products. But I'm guessing now you're doing more software stuff as well. So can we go back to your first job, Simi Robotics? What was your role there?
Shivani Mooli Swaran [00:02:09]:
Yeah, so I actually joined SIMBI as a computer vision software engineer. So at the time, I was really fascinated by image processing and specifically getting robots to see the world like people do. But what I kind of found when I was there is that it's also really, really hard to make an embedded system work functionally. Right. We were running into all of these problems where we were kind of scaling from generation one to generation two of the hardware, and things just fundamentally weren't working.
Manav [00:02:35]:
Can you explain what their product was?
Shivani Mooli Swaran [00:02:37]:
Yeah. So they do autonomous grocery retail inventory. So if you've ever been to kind of a large midwestern grocery store like Schnucks or Kroger. They are a robot that rolls around the grocery store, scans the shelves, and then tells you when products are out of stock.
Manav [00:02:52]:
So it basically runs in the back end of the store.
Shivani Mooli Swaran [00:02:55]:
It's actually in the front side. So where consumers are picking up cereal boxes off the shelf, it's actually moving through that exact same space.
Manav [00:03:03]:
Really?
Shivani Mooli Swaran [00:03:04]:
Yeah.
Manav [00:03:04]:
But doesn't that interfere with the customer shopping experience?
Shivani Mooli Swaran [00:03:08]:
That company did a really great job of building, building a really friendly looking robot that has a really small footprint. But I think watching people interact with it, you tend to see people doing much stranger things to the robot than the other way around.
Manav [00:03:24]:
That's really interesting. And then after that, how did you end up joining Tesla? Because it's no easy feat, first of all, getting a job at Tesla.
Shivani Mooli Swaran [00:03:35]:
Yeah, so I think kind of going back to, to what was happening at the startup at the time that I was there. So they were a seed stage when I joined and had kind of like a very small team of 10 to 20 people. And we were moving from the first generation hardware of the robot to the second and integrating an entirely new camera stack in at the time. And it was just a really long and painful process. I ended up owning most of that integration and discovered that, okay, I've kind of hit the limit of how deep I can go here and I really want to see how people are doing it outside of the startup at scale. A Tesla recruiter had reached out to me actually at the time and was like, hey, do you want to work on camera systems at Tesla? And it turned out to be a really good fit. I was super excited to see how this giant handles cameras on hundreds of thousands of vehicles on the road. And that's how the switch happened.
Manav [00:04:24]:
It definitely sounds like a much more complex challenge and I think. Is that the reason you took the job?
Shivani Mooli Swaran [00:04:31]:
Yeah, I just felt like I really hit kind of an upper bar on where I could go. Like there's only so much you can learn when you have, you know, less than 100 units deployed in the field versus what it looks like day to day as an engineer trying to handle hundreds of thousands of vehicles on the road that are moving in safety critical situations.
Manav [00:04:48]:
Got it. And at Tesla, how were you like, what was your exact, like, role and what kind of problem were you solving in your role?
Shivani Mooli Swaran [00:04:56]:
Yeah, so I was a software engineer at Autopilot within a sub team that was specifically focused on camera and image quality tuning. So what that means is anytime there are New sensor integrations, we kind of make sure those go smoothly. We handled a lot of the metrics to make sure that you know, when cameras are failing, why they're failing, how that translates into service if you need replacement parts, and then also the kind of long tail of, hey, we're constantly iterating on this huge AI stack. How do you make sure that the data coming in is not only coming in at the lowest latency possible, but also the highest quality for getting features out of it that then translate into, okay, can I see a pedestrian moving in front of the car fast enough that I can brake?
Manav [00:05:41]:
So I'm just going to say a few things from my knowledge and maybe you can correct me if I'm wrong. So autonomous cars are either three technologies. It's either camera based, it's either lidar or radar. So can you explain the landscape of the different companies pursuing what and what are like the downfalls of using this camera system? Because I would. This is my opinion. I would think, like in harsh weather conditions it would be sometimes really hard for the camera to detect something. And was that not your experience when you were working there?
Shivani Mooli Swaran [00:06:18]:
I think the major difference actually comes down to the, the approach and the go to market and then also the sheer cost of, you know, how much it takes to put the system on the road. So I think most people would assume that with cameras they are really, really bad in harsh weather conditions. And that doesn't apply to LiDAR. That's actually not really true. In snowy conditions, both of them are prone to kind of failure because, you know, snow reflects light and fundamentally neither LIDAR nor camera is really going to work amazingly in that situation without a lot of post processing. But I think there are unique kind of failure modes that you do run into. So with cameras, specifically when it rains and droplets land on the grass or someone smears their finger across it or some dirt is landing on it, those are all kind of like obstruction situations that you kind of have to figure out ways to get around or detect that scenario and then shut something off safely when it happens.
Manav [00:07:11]:
So in those conditions, does Tesla prompt you to just take autopilot, not go off autopilot?
Shivani Mooli Swaran [00:07:18]:
Yeah, Tesla's been really great, I think about covering a lot of those obstruction conditions and then shutting down autopilot in those kinds of situations.
Manav [00:07:26]:
Amazing. And so let's talk about how the autopilot has kind of evolved over the years. Like what was like the, when you first started and to what it is now.
Shivani Mooli Swaran [00:07:37]:
This is really interesting, I think so to be Honest, when I came into the company, I think they'd already made an insane amount of progress. I came in around 2022 and I think the growth that we've seen since then, to the release of FSD, I think 12 is just a lot more human, like behavior versus kind of like a classical controls approach. And so you're seeing the car move a lot more smoothly through these really interesting edge cases that previously it wasn't able to handle before. I think there's also been new hardware launching. Tesla in the last five years moved over to their own AI chip and that's also pushed a lot of growth in that sector of optimizing the compute on the platform so that you can push way more new features out.
Manav [00:08:26]:
That's incredible. How does Tesla do the real time processing of the camera data? That's like, how is it. Do they need some kind of Internet or is it locally processed inside the car, that data?
Shivani Mooli Swaran [00:08:43]:
Yeah, this is what's super cool, I think, about the way that Tesla's been able to handle this, and I think this is something that we are kind of excited about bringing to other people in the industry, is that they had such a really heavy focus on, hey, how much of this data can we process in real time and make sure everything happens on device so that we can make those decisions as quickly as a person is making them? So a lot of, you know, heavy signal processing techniques, a lot of smart ways of kind of. I'm trying to think of the right way to put this. Picking and choosing exactly which data is actually critical. Right. Because if you were to lift every single piece of data coming off of the device, it's just a glut of things that don't really mean anything. So picking what is actually important to analyze and then crunching that down into a signal that is usable in other parts of the system.
Manav [00:09:31]:
Yeah. Before recording the podcast, we were actually talking about how this technology one day would be maybe available in India. So can we talk about, can you talk about if that's actually. Can you make a prediction if that's going to happen or not?
Shivani Mooli Swaran [00:09:49]:
Yeah. This is a fun conversation that we've actually been having with some friends recently, is that in India, if you've been to any large city like Chennai, Delhi, etcetera, you'll notice that a lot of people aren't really. They're driving in really tightly enclosed spaces and they are honking frequently to make sure everyone around them knows exactly where they are. So vision based tech, quite frankly, at this point, isn't really the way to go. Because there's just so much noise, so much contrast and so many features floating around. We're kind of hoping that eventually someone will be able to do audio based self driving and then just echo, locate a car through that kind of traffic.
Manav [00:10:22]:
Whoa. That does not sound. Camera. Okay, that sounds. Yeah. You know, whenever I visit India, like I get triggered. So the traffic is getting better in terms of like the lane systems, but I do get triggered by the honking. It's so unnecessary. I think they need to have like some kind of ticket system for that. Like you get a ticket if you honk for no reason. I think they do actually, but I don't know if anyone's following that. But so what made you want to decide? Like, you know, like people at Tesla, like they, you know, they have this incredible, incredibly good stock option and like this stock is like rallying for years. And I mean, I know like Indian parents would guide you to just stay there and be like, hey, like, you know, you're getting stock options, you're getting the. This is a really good job. What made you want to segue into starting your own company? And was the pain point that you recognized so strong that you're like, you have to do this?
Shivani Mooli Swaran [00:11:25]:
Yeah. Two standpoints on that. One is actually interesting. My parents are indeed parents, but my dad actually worked at a ton of startups when I was growing up. So I grew up in Cupertino and he didn't really have a job at Apple. He actually just worked at a bunch of large data startups, cybersecurity, et cetera. And he did this even through the tech recession and was working on a project of his own on the side. And so I think that for both me and my brother was kind of a learning point of, hey, there's other options out there that aren't just pick a company and ride with that company for the next seven years and pick stability. Right. I think the other reason is I've been in the industry for five years and this is a problem that has persistently followed me at every company that I've been at. And I am just more optimistic about the robotics industry and where it is now, about putting out third party tooling that enables people to build a lot faster and build things that are helpful in ways that they weren't able to do five to ten years ago. Right. The cost of sensors has come down significantly. There's a lot of people building great visualization tooling now for robotics. A lot of people that are building robotic arms and separate Subassembly components. And when you put all of them together, it's just starting to get to a point where things are accelerating a lot faster than before. And I think now is an exciting time to be involved in that.
Manav [00:12:41]:
Yeah, I interviewed the founder of Chef Robotics and it's insane. They've already been deploying so many robots. And another thing that I'm noticing is like a lot of these robotics company are noticing a shortage of labor, which is a big problem because as we also have like declining population rates in the US and it's not even like oh, the robot is replacing a human, it's more like, oh my God, we don't even have people to do the job. So you're right. And people, when they think of robots, they only think humanoid robots. But can you explain what other types of robots there are in the landscape of robots?
Shivani Mooli Swaran [00:13:18]:
Oh, I have hot takes on humanoids too. So I think a lot of robotics is really focused on one single user case that is really valuable to your person or some type of multi use case platform that is really valuable. I think a great example of this is actually going back to what Symbio Robotics is doing. Grocery retail is an industry that you wouldn't think of as an entry point for really, really new cutting edge tech. But they do suffer from these huge problems where when somebody goes out on the floor and has to scan retail inventory, inventory, they're doing a very, very manual process that takes a ridiculous number of hours and they're only able to do it a couple times a week along with all the other responsibilities that they have. So they were actually able to identify the scap of. Okay, well if we put a robot in there that can run a couple times a day and scan autonomously now we can actually tell you in real time how your inventory is moving through the store and actually increase both customer satisfaction and profit margins based on that.
Manav [00:14:17]:
And that data is going back to the retailer or like who's getting that data.
Shivani Mooli Swaran [00:14:21]:
It goes to the retailer and helps them make better decisions about how to like stock shelves. But I think this is kind of the like also just.
Manav [00:14:28]:
And how is that company doing today? Do you know?
Shivani Mooli Swaran [00:14:30]:
They're doing great. They just raised it. Seriously, I'm excited to see them.
Manav [00:14:33]:
Yeah. Because I would like. Oh, that's such a great problem they're solving by the way. Because. And I would, I would imagine every store would have that, especially the big ones. Yeah, I can imagine having that and I think I've seen a video on that. So I'll Ask my editor to put the video on when you're talking about that company. Cool. So I want to, I want you to explain Ember Robotics. Like, explain it to me like I'm five. Okay. Explain me what Ember Robotics does and what problem are you solving and is it like a database of how is it a third party tool? And if I'm a robotics company, how can I use Ember Robotics?
Shivani Mooli Swaran [00:15:14]:
Yeah. So I think the fundamental, best analogy that we found for what we're doing is kind of looking at healthcare. So when you have a problem with your eyes and your vision, you go to an eye doctor, they go through a series of standardized tests, and they tell you exactly what's wrong with you, and then make a recommendation on how to fix it. When it comes to robotics and complex sensor systems, that kind of root cause analysis is a lot harder to do. And usually you spend a lot of years building up tooling in house to lift the metrics, to even make those decisions in the first place. So what we do specifically is we build real time software that goes on your device and it plugs into the sensor metrics, your logs, some of your system information, and then makes those decisions for you in an automated way. So instead of you spending all your time trying to figure out, okay, well, I know that the camera is up for this amount of time. I know that the temperature of the sensor is this much and that the CPU is x percent overloaded. And that comes to this conclusion about why it's failing. That's what we're doing.
Manav [00:16:21]:
For people right now, that sounds like a custom solution then more like, like for every robotics company is going to be like a custom solution for each robotics company. Right.
Shivani Mooli Swaran [00:16:34]:
So this is what's really fun about, I think, the landscape of sensors and robotics right now. I think something not a lot of people realize is that the same kind of cameras that you use in robotics are often the same cameras that you use in automotive. They're often the same ones that you use in drones. Sometimes they also show up in kind of traffic camera management systems. And all of them are fundamentally automatically operating on the same types of signals. Where the data comes in from is different across all of those, but the actual analysis of the metrics looks really similar across all those stacks. And so that part of it is where we're generalizing. And then the data ingestion layer is kind of something where we'll build it custom for people if they have something really unique. But a lot of people also use pretty generalized data passage protocols that make It a little easier. Easier for us to integrate with them.
Manav [00:17:21]:
Awesome. That's a very unique problem. I don't even know if anyone else is solving this, by the way. There must be some other companies. But I want to hear your hot take on humanoid robots and then I also want to hear your opinion on facial recognition because you were in the cat. I mean if every humanoid robot has a camera in it, they're recognizing these faces and we've seen like it kind of has gone sour in like China where you don't even get a parking ticket now. You just automatically they take the money from your account and then they have the social credit scoring system. But let's hear it about humanoid robots. Like, and I want to hear your prediction on where we are and where we're going to be in 20 by like in five years.
Shivani Mooli Swaran [00:18:04]:
Yeah. So humanoids, I, I don't think that form factor is the best use case for a lot of the things that you might want to solve with it. If it does get to a point where it's really, really highly generalizable and that is like the big value add, then sure, that kind of works out for some scenarios. But right now I'm more bullish on single and multi use case robotics where they are, you know, 10x more efficient than a person is.
Manav [00:18:32]:
And like a robot vacuum.
Shivani Mooli Swaran [00:18:35]:
Yeah, exactly.
Manav [00:18:36]:
Like it's just a single, like it has one purpose. What are some other examples?
Shivani Mooli Swaran [00:18:43]:
I think like mining robots are super interesting. There's a lot of agricultural robots that are again, you know, single use case. They're optimized specifically for performing in that environment and doing one specific thing about like crop.
Manav [00:18:54]:
Like the coffee machine. I made a video on that. Yeah, like the, the, it's, it's on, it's in the San Francisco airport. You know, the automatic click. It actually makes really, really good coffee. And it's crazy that you only have to stock it once a day. And it can make a cup like every two minutes. It doesn't get tired.
Shivani Mooli Swaran [00:19:14]:
Yeah.
Manav [00:19:15]:
Which is, you know, it's doing three people's job by itself. So would a company like that could be a customer for Ember Robotics. Or it has to be much bigger. Like it can be in any company.
Shivani Mooli Swaran [00:19:27]:
It could be any kind of end user. I think we're more interested in what's kind of under the hood. If they're using like a Linux based system and a specific type of sensor stack, that's kind of what we'll work with. But I think the macro take on Humanoids that I was going to give is from the perspective of the self driving industry. If you looked at how it originated, there was this huge hype cycle of guys, we're going to have fully self driving cars on every road in a decade. And that hasn't really fully materialized yet. I think a very similar thing is going to happen with humanoids. But the net positive of the self driving push is that a lot of the auxiliary technology that came out of it has kind of pushed the boundaries of driver's assistance tech. It's pushed the boundaries of, you know, how large of models can we deploy on the edge in a moving device that people are using. Right. And those are really, really have been great to push other angles in the robotics industry. So I think we're going to see see the same thing with humanoids where manipulators are going to get significantly better, you're going to get significantly better and smaller motors as a result of people trying to shove everything into the humanoid form factor. And that will have net benefits even if the end result is not an actually human looking, human functioning robot.
Manav [00:20:37]:
Yeah, I just can't wait for the humanoid robot to do my laundry. That's all I need the humanoid robot for.
Shivani Mooli Swaran [00:20:46]:
But again, you just need two arms for that. Right. What it's connected to is a different question altogether.
Manav [00:20:50]:
It's so funny. I was listening to this podcast and they're like, if you, if every person has a humanoid robot, it's going to be like, you know, you go to a restaurant and they're going to be like, no robots allowed inside. And your humanoid robot is going to be like waiting outside for you. But I do agree with you that they're going to be more single use case robots, like in the short term, at least for the next decade, because the humanoid robots like. But I totally. I have been watching the demos of Figure, the Optimus and even the ones coming out of China and they already are like, they can perform like so many tasks.
Shivani Mooli Swaran [00:21:27]:
Yeah.
Manav [00:21:27]:
Which is incredible.
Shivani Mooli Swaran [00:21:28]:
Yeah. The push is definitely important because I think without the push towards attempting humanoids, no one's going to really make those strives on innovating for all the tech that comes out of it. Anyway. That is super awesome that they're trying for it right now.
Manav [00:21:42]:
Cool. Okay, let's segue into Y Combinator. Why did you decide to go to Y Combinator? And are you done with Y Combinator right now?
Shivani Mooli Swaran [00:21:51]:
Yeah, we finished up the summer batch around the end of September earlier.
Manav [00:21:55]:
First of all, congratulations, because getting into YC is incredibly hard and you got in, so woohoo. But yeah, I want to know, like, how was your journey? Like, tell me everything. Like, how did you apply for YC and how was your journey getting in? And then how was the three months when you were there?
Shivani Mooli Swaran [00:22:14]:
Yeah. So I guess the decision to apply to the accelerator was largely because my co founder and I were bouncing around a couple of different ideas and we wanted some funding and some access to resources to kind of get those off the ground. And YC and another accelerator that we'd applied to specifically, the appeal there was, okay, you've got mentorship, you've got people that are going to walk you through the go to market process, how to do fundraising, when you get to that period, how to do sales. And both of us are first time founders, so all of that sounded really appealing. The YC interview process was, I think it was probably one of the most intense interviews we've ever taken.
Manav [00:22:57]:
Really?
Shivani Mooli Swaran [00:22:57]:
It's only 10 minutes long and they just cram it full of all kinds of questions. It kind of feels like you.
Manav [00:23:02]:
Was it in person or over zoom?
Shivani Mooli Swaran [00:23:05]:
We actually did ours in person.
Manav [00:23:06]:
Okay.
Shivani Mooli Swaran [00:23:07]:
They had been doing them over zoom, I think, until our batch because of COVID and kind of coming out of that.
Manav [00:23:12]:
And how was it?
Shivani Mooli Swaran [00:23:14]:
Really intense. Yeah, you just get peppered by a lot of different questions. They explore everything. Every, you know, soft spot in your company plan or how you think your company going to go to market, they will find it and they will go down that path and you have to think of, you know, what your response to that kind of soft point actually is on the spot and deliver it.
Manav [00:23:36]:
And who was doing the talking? It was you or your co founder.
Shivani Mooli Swaran [00:23:40]:
Both of us actually.
Manav [00:23:41]:
Both of you guys. Nice. Okay, so you got into YC and then I'm guessing it's like a three months intense boot camp where, where you literally like, you know, you know, by the way, I learned this China has this. China had this principle of996 where you work nine hours every day and then on Sunday you work six hours. But there's this new rule called 007 where basically if you're an executive in China, you work all seven days, all the time, and you have no other things to do. So how was it when you were in yc? Like, what kind of value did you get? Like just being around that environment, Were you showing up to their headquarters every day? Like, how was it?
Shivani Mooli Swaran [00:24:23]:
Yeah, I think, I mean it definitely is a lot of hours, but I think we were both really used to that coming out of Tesla and then working at a startup prior to that, the hours were relatively similar. I would say the thing that was really intense and very different for both me and my co founder was that you have touch points every two weeks with your mentor and kind of the goals that you're trying to reach in that time period are always high and you're trying to get them done really, really quickly. So often this comes in the form of, hey, I have this goal of I'm going to talk to X number of people this week to validate that this idea is actually a problem for people, that it is a pain point they want solved, and then start building small pieces of software to experiment with. You know, is this actually the thing that is going to solve the need that they have and then scaling into, trying to do contracts, acquire customers, et cetera. And by the time you realize that you've already hit the end of the batch and now you're in fundraising mode, and that's kind of another intense period in and of itself.
Manav [00:25:21]:
Yeah. I'm curious how the demo day was, because that must have been like nerve wracking a bit to present at the demo day.
Shivani Mooli Swaran [00:25:29]:
Yeah. So weirdly enough, by that point in time, I think we had already dealt with kind of so much in the growth of the company in that three months. That demod itself was not the most nerve wracking piece.
Manav [00:25:40]:
Got it.
Shivani Mooli Swaran [00:25:41]:
I think we were having a lot more fears around. Okay, well, we're kind of exiting the batch now and we had gone through it with this community of awesome founders where part of the value in YC was being able to move through the batch with them and watch as they were growing their companies at the same pace and kind of exiting it and not seeing them on a really regular schedule because of YC was a huge shift.
Manav [00:26:05]:
Yeah. Did you maintain that same work schedule after? Probably not.
Shivani Mooli Swaran [00:26:12]:
Yes and no. I think we both had a little bit of a settling period. My co founder went to visit some family and then I was actually moving up to the city. And so there was a period of time where I think I spent a week just trying to figure out how to settle down. Junior apartment.
Manav [00:26:27]:
Awesome. Cool. I want to kind of segue back into Ember Robotics and talk about like, fundraising. What do you guys are focused on right now? What's the next big step for Ember Robotics?
Shivani Mooli Swaran [00:26:38]:
Yeah, so we close around on fundraising, kind of focus back on building and kind of, I think both building the product for people that we're working with currently and then also exploring new avenues of where the product really should go. I think it's been interesting recently because we've built some new features into the product that aren't just tackling. Okay, here's your base one of reliability, which is, does my hardware kind of work and what are the reasons that it's failing that are just coming out of the actual system integration itself. And it's now coming into people being a little bit more interested in. Okay, well, once I have those in place, how do I tune my sensor so that they're actually performing in a way that I can get useful data out of it or that I can deploy some model onto the edge? So we've been building some stuff around image quality analysis, time synchronization, error detection, and things in that area to try to test and see if this is going to solve those issues.
Manav [00:27:35]:
Got it. And can you talk a little bit about image recognition? And I know companies like Scale AI, they have an army of people in countries like Africa and India whose entire job is to just like identify what an image is. And that goes back into training the model. Can you talk a little bit about that? And is that data publicly available or how do you get that data?
Shivani Mooli Swaran [00:28:03]:
Okay, so we don't sit specifically in that loop. So the image recognition aspect, that usually goes into people that are trained to do specific object detection or they're trying to get their model to detect a person or something like that, we set a layer below that. And so we're actually not even consuming full images at this point. We're consuming the metadata that is stashed in the images itself. That tells you a lot of the exposure values, kind of like what the image looks like in terms of just raw settings. And that's how we're able to make those judgments. Versus, like, okay, I'm going to go through a bunch of 4K images coming in and then try to figure out how to do those really quickly and then make those judgments based off of it.
Manav [00:28:49]:
Interesting. And how would you like, do you. This is not currently deployed, right? Or is it. Have you experimented this on any product yet?
Shivani Mooli Swaran [00:29:00]:
The image quality analysis itself is not deployed into production yet.
Manav [00:29:04]:
Okay, amazing. Yeah. I'm really stoked to do a recap in a couple years and see what kind of companies you've implemented this on. And how would you like, what's your strategy? I mean, you don't have to be too specific, but just to give other people ideas as well. What's the go to market for this kind of products? Like, are you just essentially going to be reaching out to other robotics company. And what did YC advise you to do?
Shivani Mooli Swaran [00:29:37]:
Yeah, the way that we've thought about this and kind of the way that we've kept going with this through YC is fundamentally the problem that we're trying to solve is that for people that are engineers like we were in the same position at our previous jobs, they're just wasting way too much time banging their heads into the wall trying to figure out why something is going wrong when they really should be focusing on building core product features that are serving their end customer. Right. I never want to be the person that has burned half their week goose chasing some kind of hardware issue when I could have been building better navigation or better data analytics that then tells them something important about their environment. Those are the people that we're trying to solve for. And the way that we're kind of going to the market right now is working primarily with robotic startups. So that could be anyone from drones to kind of like mobile robotics and warehouses to the grocery robots that I mentioned before. And for them, what we're trying to do is just help them build a lot faster and scale a lot faster. Because historically software has kind of hit the point where I can go into some random tool and spit out a bunch of components and throw up a website really, really easily now. And that was really, really hard to do back when everything was just raw HTML. Hardware has not built up enough tooling around it where you can accelerate it to that same point. And that's kind of the end goal here, is we just want to take it, you know, two times faster, three times faster, make sure that it doesn't take a decade to put a robot out because a lot of companies will just end up going under in that time frame anyway.
Manav [00:31:08]:
Yeah, and that's what happened with, with the drone industry. Like Biden administration tried to kill the drone industry and now we're having this problem with the biggest drone manufacturers, dji, which is the company based out of China. And it kind of is a security risk to us at this point. And I mean, yeah, like I think regulation is one of the main reasons again like why we don't have enough like self driving cars. But I, I guess you got to play that tug of war. Okay, now I want to ask you some personal questions. Segue out of your company and get to know more Shivani, the founder, because you're super interesting. I want to know like, how do you unwind? Like what do you do for fun? What do you like? Okay, by the way, Guys, Shivani plays sitar. So she is an. Sitar is like an Indian. What would you say, violin or Indian guitar? It's like a really long Indian guitar, but it's beautiful. It sounds amazing. Can you tell me a little bit about your experience playing a musical instrument?
Shivani Mooli Swaran [00:32:14]:
Yeah. So my mom actually grew up playing Siddharth in India, which is how my family got into it. And I've been playing since I was around seven. So I grew up taking lessons playing Hindsani classical music until I went to college, and then have been kind of like periodically playing on and off ever since. I actually have a Siddharth teacher in India who I will take zoom classes with still. But I think what's really cool about Siddharth specifically is that if you are an engineer, it's actually really fascinating from a mathematical standpoint because it's very improvisation based. And a lot of the pieces that you play are meant to fit into a very specific beat cycle. And then within that beat cycle, you basically do fractions to figure out how you can fit the notes into it to build a cohesive melody.
Manav [00:32:57]:
Jesus.
Shivani Mooli Swaran [00:33:00]:
I think it's just a.
Manav [00:33:02]:
One second.
Shivani Mooli Swaran [00:33:12]:
It heard about the math patterns and just gave up.
Manav [00:33:16]:
I was like, damn, what do you do for fun?
Shivani Mooli Swaran [00:33:20]:
We do math.
Manav [00:33:21]:
That could be a short. Like, that could be like a reel. Where I ask you about sitar, and you gave me such computational nerdy answer. That's. That's actually hilarious. It's okay. I'll just record it on my. Because we don't have any space. Oh, sorry. Try to record on my computer. I'm gonna have to do this, like, idiom 4K. Oh, yeah. This actually looks. Doesn't look bad. It's 0.5. This doesn't look bad at all. Okay, that's good. Okay, cool. So I have, like five questions I ask every guest on the show. And you don't have to give a very long answer. You can just give. Be brief. But the first question I ask everyone is because I love reading, and I have read probably 100 books in the last two years. I want to ask, what is the book that you've given to other people as a gift, and why? And if you're not the person who gift people books, what are like one to three books that have influenced your thinking and really shaped your life?
Shivani Mooli Swaran [00:34:51]:
Oh, this is great. I actually grew up reading, and I really, really love reading. Longitude by Dava Sibyl is a fantastic book. So it's basically the history of how timepieces were invented to solve the problem of people getting lost at sea because historically you could calculate latitude really, really easily by kind of the position of the North Star or just kind of knowing, I think, the angle of the sun towards your ship at any given point in time in the ocean. But in order to know longitude, you actually needed to know the time differential between two different clocks. And they only had pendulum clocks at that time. So at sea, obviously those are oscillating all the time and everything's completely inaccurate. So people would look, literally go out to sea to go somewhere, get completely lost on the longitude, and then just end up in the middle of nowhere and pass away. Right. And so there was a huge race basically between a bunch of European countries where they put up a. Not a ransom, but a reward for anybody that kind of solved this issue. And there was this guy who came in and invented basically the first early analog clocks that the then kind of became the timepieces and the watches that we use today.
Manav [00:36:00]:
Interesting. Wow. And what year was that? Is it 1800s or even before? Probably before, right?
Shivani Mooli Swaran [00:36:09]:
Yeah, I want to say the 1800s, but don't take that with a grain of salt. I'm bad with remembering.
Manav [00:36:13]:
Got it. Was that the only book you recommend others to read?
Shivani Mooli Swaran [00:36:16]:
I think Demian by Herman Hess is also a really good one. So this is kind of a coming of age story and it's an exploration of basically the identity crisis that a guy goes through where he's struggling with societal expectations versus what he wants from the standpoint of, okay, what does it mean to really be a spiritual person? What does it mean to really reach a higher plane of thinking as a human being that has to deal with the pragmatism of everyday life? And it's set around the time of one of the world wars, so there's a lot of that kind of conflict of.
Manav [00:36:46]:
Go ahead.
Shivani Mooli Swaran [00:36:47]:
I think humanity versus just being involved in a large scale world conflict. That comes into it too.
Manav [00:36:53]:
That was actually. Oh, I have to say that was the best response I've got on this podcast so far. That was so good.
Shivani Mooli Swaran [00:37:01]:
What books do you normally hear recommended?
Manav [00:37:03]:
No, it's not even about the book, but you just gave such a good explanation of both books. Oh, my God. You're like, you know when you ask ChatGPT to summarize a book. Yeah, that's what you just did.
Shivani Mooli Swaran [00:37:17]:
That I haven't actually been able to finish recently because I just haven't had the time. Yeah, it's literally called Time Management for Mortals. And I haven't succeeded in getting More than two chapters in because I haven't been able to manage my time.
Manav [00:37:29]:
The irony of that. Okay, next question. What purchase of $100 or less has most positively impacted your life in the last six months? It could be anything. It could be a book, course, movie, it could be a gadget. It could be anything you bought on Amazon.
Shivani Mooli Swaran [00:37:46]:
This is just one that I've been buying since high school. 10 to $15. Casio digital watches are an absolute workhorse. I actually have been running since I was really small, and I don't really use Garmins or Apple watches. When I run, I will literally just turn on the stopwatch on my Casio and just keep going with that. And they just last a ridiculously long time.
Manav [00:38:04]:
They cost $15.
Shivani Mooli Swaran [00:38:06]:
The really basic one that's insanely low plastic, that all it does is it counts the time, it lets you do a stopwatch, and then it tells you what day it is.
Manav [00:38:16]:
I love that. You know, I've been against the Apple watches. Not that I don't love the tech. I just feel like it's another distraction on your wrist and I'm like, really trying to get away from screens. So I would actually be a customer for this Casio watch. Cool. Question number three. How has a failure or apparent failure set you up for later success? Do you have a favorite failure of yours?
Shivani Mooli Swaran [00:38:42]:
This one? I think, you know, maybe not everyone would consider this a failure, but I think the. The area that I grew up in, the school district that I came out of was very academics focused. And so people were kind of constantly competing on grades, competing on extracurriculars, competing on where they went to college. And I think coming out of that environment and going into college, I actually ended up choosing to do a lot of side projects just for fun. So I was doing drones with ieee. It was kind of like messing around with a lot of different projects with circuits, etc. In the lab. My GPA definitely suffered as a result of that. And it was kind of borderline for a while in college, which didn't feel so great because it did kind of feel at the time with that kind of environment that, oh, maybe you're not going to succeed, you're not going to get a good job because of it. But it was because of those side effects projects that I was doing that I was actually able to get into robotics and find something that I really, really liked. Because, you know, in robotics, your academics only matter so much as what you can actually apply in the real world. What can you actually put out there physically that's doing something meaningful.
Manav [00:39:45]:
Yeah, no, I totally agree. Like I was never good at school, but I consider I turned out fine, so. But you're right, it does affect you mentally to like not have good grades. You kind of builds that self doubt, you know, like, oh, am I not that smart? But the extracurriculars come in handy later on in life. Okay, next question. What is the one? What is one of the best worthwhile investments you've ever made? It could be an investment of money, time or energy. It could literally be like you invested in some stock like Nvidia probably six years ago. It could be anything.
Shivani Mooli Swaran [00:40:25]:
I participated in GameStop. I don't know if I would say this was the most worthwhile, but it was the most entertaining investment I've ever had.
Manav [00:40:34]:
Awesome. Were you not stuck in the Robinhood scenario when they stopped buying?
Shivani Mooli Swaran [00:40:39]:
Well, I was on there, so what happened with me is that I bought some options and I was kind of watching them spiral and I sold pieces of it off so I wasn't completely upset about it. But I did watch that whole thing go down and I think it was, it was a fascinating episode of kind of learning how the markets work and all the regulations around it.
Manav [00:40:54]:
Yeah, it's funny, they were allowing to sell but not to buy. How is the system not rigged? Okay, last question. In the last five years, what new belief, behavior or habit has most improved your life? It could be anything. It could be fitness related, health related, psychologically a behavior you changed or it could be something you learned at yc. Could be anything I think just generically.
Shivani Mooli Swaran [00:41:23]:
Taking action on a lot of things and not really overthinking it. I think when I was in high school I was really a much more practical like step by step type of person. And then after going to college and kind of working at startups, etc, I started becoming a lot more okay, hey, if I want to do something, I'm just going to apply for it. Doesn't really have matter what the result is. If I want to talk to someone, I'm actually just going to cold outreach to them and ask them to grab coffee because I'm interested in hearing what they want to say. And I think that actually is one of the reasons that I'm able to sit here and say that I founded a startup is because your risk aversion kind of decreases over time because you realize nothing is ever really as bad as it seems in your head in the aftermath. Even if you get rejected by trying for something.
Manav [00:42:05]:
Yeah, I tell everyone to take more asymmetric bets in life. And I think we have to unlearn a lot of things we've learned in school. And I think this is my opinion, but I think school makes us, like, slows us down, because in the real world, you have to move really incredibly fast and take a lot of actions. But it's been great to have you on the show. How do you feel? Should I ask you more questions or is that it?
Shivani Mooli Swaran [00:42:34]:
Yeah, this has been super fun. Really appreciate you having me.
Manav [00:42:37]:
Okay, so thank you, everyone, for listening. That was Shivani from Ember Robotics, by the way. Shivani, like, how can people get a hold of you?
Shivani Mooli Swaran [00:42:47]:
Yeah, so you can reach out to us through the demo page on ember robotics.com. we're really active on LinkedIn if you want to look us up there. Unfortunately, don't have a Twitter account that you can can take over right now because somebody's currently posing as us on there. So don't post to that Twitter account.
Manav [00:43:04]:
Are you on LinkedIn?
Shivani Mooli Swaran [00:43:05]:
Yes.
Manav [00:43:06]:
Okay, Shivani's on LinkedIn. And guys, you spell it. E M, B E R Robotics dot com. Make sure to check them out. They're going to be building something really cool and we'll do a recap episode in a couple years to see where the journey has led them. Thank you for listening. That's another episode of Emerging Founders with Manav. I'll see you in the next one.
Related episode
