83: NaadSadhana – Sandeep Ranade

Sandeep Ranade:

I've had offers to buy out the app and the only question I ask them every time is, "What feature would you add next?" They had no answer. It now truly seems that I'm the only person who's passionate enough to evolve this app in the way that I care about, in the sense that an artist, a performing artist cares about.

Charlie Chapman:

Welcome to Launched. I'm Charlie Chapman, and today I'm excited to bring you the developer behind the Apple Design Award-winning music app, NaadSadhana, Sandeep Ranade. Sandeep, welcome to the show.

Sandeep Ranade:

Thank you for having me on your show. It's such a pleasure.

Charlie Chapman:

Yeah, this is going to be really fun. It actually wasn't very long ago that we met for the first time, and we did a live version of Launched. It wasn't recorded, and it was with Raja who makes Lumi and Kalzi and a bunch of other apps. So, it was more of a panel kind of thing. And after that, I was like, "We have to get you on the show proper and talk about your app and your experience," because it's super cool. So, I'm super pumped about this.

Sandeep Ranade:

I'm very excited. I had a great time talking to you that day, and I wanted to continue the conversation, so this is perfect.

Charlie Chapman:

And watchers of the video version of the podcast might notice that I'm in a dark room right now, and that is because you're coming live from somewhere in India. So, hopefully I will have the energy level that you've come to expect from Launched. Not that this is this intense podcast. Maybe you'll get a loopy version of me. I'm not really sure, but it'll be fun no matter what.

Sandeep Ranade:

Unplugged Charlie.

Charlie Chapman:

Exactly, exactly. But before we get into unplugged, Charlie, I want to introduce everybody to who you are. So, the questions I always ask everybody to kick off this show is, where are you from? Do you have a formal education related to what you do? And then let's talk about what your career was and what led you up to NaadSadhana.

Sandeep Ranade:

Sure. So, I'm from Pune, which is a city close to Mumbai in India. And I do have a formal education in computer science. So, I did my bachelor's of computer science here in Pune. And then I went to do a master's in computer science from Johns Hopkins University in America in Baltimore. I've worked at Microsoft. I've worked at several startups, then at Google. And so, I have a career in the US that spanned about 12 years. And then I moved back to India. I quit my job and my wife was at Apple. She quit her job. We moved back to India. It'll be 11 years in January that we've moved back. And so, we moved back to do something that we were interested in that we wanted to do.

And we started a couple of ventures and then came the two biggest startups of our lives, which is our twin boy and girl. And then the app started a couple of days after they were born. So, it was like having three kids at once. Oh my goodness.

Charlie Chapman:

You said you started a couple of ventures before that. You came back and you were trying to start a startup of your own?

Sandeep Ranade:

Yes. So, I was always fascinated by the interview process, especially at Google and the way that they...

Charlie Chapman:

Well, yeah. The interview process at Google was famous for being wild and weird questions and stuff like that.

Sandeep Ranade:

Exactly. But I was more interested in the people that didn't make the cut and why. Because everybody that applied on paper looked amazing, but it was the little things that sometimes threw them off or they went into a rabbit hole and they couldn't get themselves out of it or there were some personality issues or ego issues. And so, I was trying to figure out, is there a way that these things can be changed, can be learned and practiced and changed so that they can get into Google or Stanford Harbor, wherever they want to go. And so, that began a journey of discovering and figuring out what the most important skillset is. And I realized that this has never been taught to me formally.

And many of them, I had to learn the hard way. So, I called the company endless evolution because that is what I wanted it to be. And there were things like how to be creative on demand, how to make good decisions while still having ambiguity in your data, or how to work well in a team, how to be poised under stress and in difficult situations, how to be loved, how to be charismatic. So, all of the things that people tell us that these are the most important things for success, but are never really formally taught. So, I made small algorithms that can be practiced individually and then put together.

And some of them are biohacks, some of them are neurological hacks, some of them are just practicing techniques over and over again until they become second nature. And so, that I started when I came back. I took a year off because it has been a long, hard-working career, so to speak. And so, I took a year off. It took that time to design and build my house here and then idea started coming in and that I did for a couple of years.

Charlie Chapman:

Was this like a course then, or was this like an app or a website? What are we talking about?

Sandeep Ranade:

It was more corporate training. So, we did this for smaller to larger companies and either their engineering teams or their corporate team, their executive teams or everybody combined.

Charlie Chapman:

So, companies would bring you in?

Sandeep Ranade:

Yes. It would be like an offsite event, like a day off and then we would go to some resort somewhere and then we would train them or we could go onto their company premises and then hold these trainings. It was part of it was instruction, part of it was practice. A large part of it was practice.

Charlie Chapman:

And now this was a departure from what you were doing at all these companies, right? You were writing code?

Sandeep Ranade:

So, I started with kernel programming. I was in the Windows Kernel team, the most veteran team in Microsoft. And it was a very interesting learning experience.

Charlie Chapman:

I bet.

Sandeep Ranade:

I did operating systems, distributed systems, storage, file systems. So, all the low level of systems level stuff. Then I was doing large data pipelines, analytics, a lot of map produce kind of stuff. So, I did deduplication, we did a storage appliance at a startup that I was doing. So, it was a lot of different things.

Charlie Chapman:

Now, none of those sound like either a consumer app nor music based.

Sandeep Ranade:

So, it all started in 2017, I think the later part of the year, because there was a student of music. So, before I get into that, I'm also a Hindustani classical singer and I've been training since I was very young, perhaps six years of age.

Charlie Chapman:

Can you explain what that means?

Sandeep Ranade:

So, Hindustani classical music is a genre which is the north of India and because there is Carnatic classical music, which is the south of India, but it is all classical music and this is raga based. So, there is a lot of improvisation, a lot of spontaneity. Much of it is created on the spot. So, live performances are all spontaneous. And so, I studied, my guru was Padma Vibhushan, Sangeet Martand, Pandit Jasraj. And I've been doing that for a while. And so, I was also teaching music for a while, since 2005 actually.

Charlie Chapman:

So, you were teaching that on the side as a fun hobby kind of thing?

Sandeep Ranade:

Exactly. And so, there was a student of music who had seen me live in concert and she said, "Please help me. I'm frustrated with my journey. I'm not making any progress." And she had learned for eight, nine years from other people before that. So, I said, "Sure, let's see what I can help you with." In the first 10 minutes, it became painfully clear that she had major problems in note recognition itself. Now you wouldn't expect this from somebody who has learned classical music for nine years. So, this was like a huge shock to me. My first attempt at debugging this was to get her a tone deafness test.

So, I asked her to go to a doctor because if you're tone-deaf, you can't distinguish between two frequencies. Very, very rare thing. Most people are not tone-deaf. But if that is the case, then there is no possibility that I can teach music or she can learn.

Charlie Chapman:

And so, the issue was like she couldn't hit the right notes or even recognize when she was hitting the notes?

Sandeep Ranade:

There are 12 points on the octave, right?

Charlie Chapman:

Mm-hmm.

Sandeep Ranade:

So, she couldn't recognize half of them. She would just know the big ones and then all the minor notes she would miss out on, just not recognize. So, it turned out that she was not tone-deaf. So, this was even more of a conundrum because now how do you debug this? And so, then digging deeper, I realized that basically she had gotten into some bad habits. So, her neural pathways were a little bit messed up. And what that meant was that when she was learning and practicing, she didn't get enough guidance to course correct as she was making the mistakes. So, those mistakes became the habit. So, now I was trying to figure out how to fix this.

And the obvious answer is to give her guidance whenever she practiced, but I couldn't do that because I mean, I had a family, she was in the US, there's a timezone difference. I couldn't teach her for more than an hour a week and then certainly not give her guidance whenever she practiced. So, I thought, "Hey, it's 2017. There must be an app that does this by now." So, I just looked at the app store hoping that somebody had already solved the problem. But sadly, there was no single app that did this for human vocals. There were many guitar tuners.

Charlie Chapman:

Yeah, yeah. That's like a pretty common category, right?

Sandeep Ranade:

Exactly. Guitar tuners were thousands of guitar tuners on the app store, but not a single app that did human vocal tuning and guidance.

Charlie Chapman:

What's different about that? The notes are the same, right?

Sandeep Ranade:

That's what I thought till I went deep into digital singular processing. And there were a lot of interesting tuning algorithms that had come out in the last three, four years before I started. And there was one in particular which promised to be very, very accurate, very fast. So, I basically, without really understanding how it did its stuff, I implemented it in Swift.

Charlie Chapman:

Had you written Swift before?

Sandeep Ranade:

All of these things I was learning for the first time, just to solve a problem for her. But I thought I'm a software engineer, so I should be able to pick this up and do some damage with it. So, I started looking into tuning algorithms and this one in particular, it worked very well. I didn't understand what it was doing because I was just looking at its pseudo-code and just put it in practice. It took my iPhone 7 plus battery from about two days to 20 minutes. I could have probably made an omelet on my phone. It was that hot.

Charlie Chapman:

Goodness.

Sandeep Ranade:

So, now I was like, oh, I can't use this directly. I have to understand how this is working and all of that. So, now it was a full deep dive. And digital signal processing is something that I learned at a very shallow level in my bachelor's days, right? And I realized how shallow it was really and how deep it actually goes and the rabbit hole goes really, really deep and it was all news. I love learning new things. So, it was fun for me. And because I was not on a deadline, I was not on a deliverable. So, I looked into a topic called psychoacoustics, which was even more fascinating because this is the psychology of how humans perceive sound.

And so, there are things like just noticeable difference, how the ear actually maps different frequencies to different regions of the inner ear and how that works and how resolutions and accuracies change across ages, across different people, different frequency ranges. So, it was all fascinating.

Charlie Chapman:

Just for my understanding, it sounds like your goal at this stage was you wanted a person to vocalize a note and for it to be able to show visually how close to hitting that note it was, right?

Sandeep Ranade:

Exactly. What note it was and how close is it to the standard. Now the problem is that with human voices, with instruments, typically the fundamental frequency in the FFT is very, very strong. So, the amplitude of that, fundamental frequency is very detectable and you can just bucket it and then pick out the largest sized bucket with human voices or with complex instruments like the violin, for example, or the saran gear, it's not the case. The fundamental frequency might be actually lower amplitude than some of the other harmonics. And this is what gives the voice its unique timber, but it also makes the analysis very difficult.

Charlie Chapman:

Apologies for very ignorantly going a little deep on this, but it's because a human voice is actually producing multiple frequencies at the same time and...

Sandeep Ranade:

Any instrument will do that, but...

Charlie Chapman:

I've never thought about this before, but that makes sense.

Sandeep Ranade:

And that is what gives each instrument its unique, recognizable timber as it is called, right? What differentiates a guitar sound from a sitar sound, from a trumpet sound. They're all sounds, but the mix, the flavors, so to speak, of the harmonics is what gives it its unique character.

Charlie Chapman:

So, more complex instruments, which the human voice is in that category, it's more difficult to pick out what the correct frequency is that the person is hitting.

Sandeep Ranade:

Exactly.

Charlie Chapman:

What did you call it? Fundamental frequency?

Sandeep Ranade:

Fundamental frequency. That is the pitch that it's supposed to be at.

Charlie Chapman:

Yeah. Or that's what's perceived as the main frequency, I guess.

Sandeep Ranade:

And that perception is where acoustics come in. Frequency is objective, but pitch is perception.

Charlie Chapman:

Man, you're giving me a lot of Wikipedia articles to go in a rabbit hole in after this, but okay, so we'll keep moving forward. So, you got way deep into this rabbit hole, obviously, but then what came out of the other end of that?

Sandeep Ranade:

So, it took me nine attempts to finally get an algorithm that was correct, that was accurate, that was fast, and that was battery efficient. And then came side pieces of the puzzle, which were things like, so if you have a very noisy environment, how do you still pick out the note correctly? So, instead of going the noise oppression way, because noise suppression tends to remove a lot of frequencies that are actually useful in the analysis. So, I looked at it as music detection. This was a different way of solving the problem and it worked really, really well. I was quite surprised actually.

And so, with these two, three pieces, and the sacker acoustics also comes in how I objectively define the level of the tuner, so how difficult it is, how easy it is. So, plus minus what number of scents defines the easy level, defines the advanced level, defines the expert level, was a perceptual problem. So, with that piece, I put on a very simple user interface, which was basically just a circle that showed a left and right arc. So, if the note is flat or low, it will go left and become blue. If it is sharp or high, it would go right and it would become orange. And if you hit it correctly, it would become the whole thing would glow green.

And then if you held it, sustained it correctly at that frequency without deviation, it would suddenly become a fat green, which was very satisfying. So, a little bit of dopamine there to reinforce that habit of that perfect note.

Charlie Chapman:

Okay. So, you built that. Now at this stage, you're literally just building this for your one student and also your own fun?

Sandeep Ranade:

I wanted to try out the thing on myself and see how far I can go in the accuracy axis. And so, I built a super expert level just for myself, which was supposedly less than the just noticeable difference of the ear. So, it was under the resolution of the human ear. And it took me about two months of daily practice, but I was able to cross into that through other means. So, I was able to triangulate with the reference drone and... I don't know how it actually works, but it worked really well. So, this was for me and for my student, one student.

So, I sent her a beta version to see how she would respond to this, how whenever she practiced, if it gave her guidance, the theory was that her neural pathways should realign and come back on track. I was surprised at how quickly she was able to go from a score of 5% to 80%. It took her two weeks. I was not expecting this. She was doing about an hour a day for two weeks. So, this was mind-blowing to me. She was, of course, ecstatic because now her music could finally progress.

Charlie Chapman:

Well, and also to be clear, she was already, before this, she was spending an hour a day practicing or something, right?

Sandeep Ranade:

More inconsistently than that, because she had been doing other things. She was frustrated.

Charlie Chapman:

Maybe it was more motivating to be consistent, but part of it is also the fact that it's...

Sandeep Ranade:

No, it wasn't just consistent practice because before I gave her the app, she was learning for about two months and I was trying to get her to practice every day and she did and that didn't help because the missing piece was the guidance.

Charlie Chapman:

Yeah. It's like having a person in the room who can tell you-

Sandeep Ranade:

Exactly.

Charlie Chapman:

... immediately if you're on or not.

Sandeep Ranade:

And this is where there was an ancient system of learning in India called the Guru Cool. Guru is the teacher Cool is the home. So, the student would go live with the guru traditionally and learn the craft for years and would be instructed and guided for eight, 10 hours a day or more. And this created truly the golden age of Indian classical music, in my opinion, because there's no way that a student would go wrong in this system. Couple of weeks, couple of months, and your neural pathways are super highly precise and will never go wrong after that. But today, students are learning an hour a week practicing by themselves many times a week without any guidance.

And the danger is that if you practice something multiple times without knowing whether you have gotten it right or not, that is where things start to become dangerous. So, that is a problem I wanted to solve. And then with one student who was showing remarkable progress from a very disadvantaged score to begin with, I thought maybe it'll help five other people. So, I decided to name it NaadSadhana. Nad is the essence of music of sound itself. And Sadhana is systematic dedicated practice, devoted practice. So, I thought it's a fitting name for what the app is trying to do and I released it on the app store. So, this was how my first release happened.

Charlie Chapman:

Did you do anything to promote it? Did you have more beta testers than just your student at that point? Or was it literally like, let's take this thing that I made and just put it on the app store, give it an icon, gave it a name and see what happens?

Sandeep Ranade:

That's basically what I did. I just put it up on the app store and I gave it as beta tests to my students, but that was it. I never realized that an app needs to be promoted and there's marketing involved and all of that. But again, it was such a simple app that I thought maybe I will wait a little bit and add some more meatier features before I think about promoting the app. Because to me, it was just a single tool right now. It wasn't really an app. And so, that was a second part of the second set of features that I decided to add. And this was...

Charlie Chapman:

Well, before that, how did that launch itself go? Was it what you were thinking? Not really many people downloaded it because people didn't really discover it?

Sandeep Ranade:

It was just word of mouth. Some people said to other people that, "Hey, I've been using this app. My notes are improving." I told it to a few people that I know and that was it. I probably broke even my developer fees. I was able to recover that, maybe 2X, 3X of that. So, that was it. I never intended to make money with it when I launched it because I wanted her to get good at notes. Because if my students are very good at their notes, then it saves me a lot of time. Rather than correcting notes, I can focus on teaching them more advanced, more fun topics. So, that was a win for me.

Charlie Chapman:

You said you made some money. So, it had some sort of payment in there initially. What was the business model?

Sandeep Ranade:

It was an upfront payment of I don't remember what it was. It was not small. It was not like a dollar. I think it was like $10 or something. I don't know why I priced it. I just didn't...

Charlie Chapman:

Yeah. It sounds like at this stage it wasn't like this is my new business and livelihood. It was like the cool thing about the app store is you can take something you made even for a small personal reason and you could publish it to the entire world and see what happened.

Sandeep Ranade:

Exactly.

Charlie Chapman:

From there, you then started building on top of it. What was the motivation behind adding more?

Sandeep Ranade:

So, for that, we have to go back a few years and there have been apps and there have been these machines for accompanying specifically for classical music, but now you can extend that to all genres. But there have been drones and there have been drum or tabla kind of machines, which basically play a unchanging loop, like a metronome. It's music, but it doesn't change. So, it's great for initial practice, but as soon as you start to perform a little bit more spontaneously, that becomes a barrier.

Charlie Chapman:

Yeah, because that's something that's unique about this style of music, it sounds like, is... I mean, there's other styles of music like this, but it is this organic, different every single time experience. So, you wouldn't necessarily want just a prerecorded song to perform to.

Sandeep Ranade:

No, it's like jazz, right? If you have a drummer playing exactly the same thing every time in jazz, it'll still be the same problem. Right. So, I thought now that my app has years, what if I give it a brain? And these are all shots in the dark. I had no idea what I was getting myself into. I'd never done artificial intelligence before. I had dabbled in the theory of genetic algorithms and evolutionary program design and given some talks on it at a student level because it fascinated me. And I'd taken AI course at my master's level.

Charlie Chapman:

And remind me, what year-ish are we talking about? Because obviously AI is something very specific now that is not the case, not that long ago from now.

Sandeep Ranade:

No, this was probably a few months after, so mid 2018.

Charlie Chapman:

Okay. So, this is even pre-pandemic, so yeah.

Sandeep Ranade:

Yeah, yeah, yeah. And this is before LLMs and all of that. Yeah. So, AI wasn't really a commodity at that point in time, and it was less accessible, but all of the problems that I wanted to solve had never been done before. And so, things like creativity in a melodic sense, creativity in a rhythmic sense, appropriateness, aesthetics, what is beautiful, what is not. All of this was never really done. And so, I started inventing solutions, homegrown solutions sitting in my little home office here.

Charlie Chapman:

And solutions, meaning your goal at this point was you specifically wanted something that would synthesize music that was organic and ever-changing and everything, but it was playing the music itself from the app?

Sandeep Ranade:

Yes, it was creating and playing music in the real time spontaneously. And so, I started with an instrument which is called a Swarmandal, which is basically an auto harp, but without the keys on top. So, just a harp 40 strings, and you tune it to a particular scale or rag, and it takes a long time to accurately tune 40 strings, but the effect is magical. It sounds fantastic. These are plucked. That is more auto harp. You specifically play it. Here, you will basically strum it in patterns, but you can also play it. You need long fingernails for it. So, there were logistics like traveling with this instrument became a pain, especially in the later years because airlines started becoming very interesting.

And then audiences also don't really have the patience for me to sit and tune the instrument for 20, 30 minutes every time I change my scale or rag. So, I thought this is a perfect use case for something that is a portable instrument, it is in my phone, I can pull it out at any point in time and play it and it'll play intelligently, creatively, artistically, spontaneously, and recreate the magic that I experience when I play the Swarmandal myself ideally. And this was a hugely ambitious goal because I never thought it would even be remotely possible, but I wanted to give it a proper shot. And this is when I started delivering into AI, realizing that existing techniques don't work for me.

So, I started inventing techniques. I have a fascination with neuroscience and learning theory and biohacks and all of these topics. I also have fascination with other topics like mathematical topics, so graph theory, discreet mathematics, optimization theory. And so, somewhere I was able to combine some of these ideas and something came out of it which was surprisingly human-like at that point. And so, I thought that, yes, this is actually promising. I can actually go deeper in this. And so, I implemented the Swarmandal. There were problems with the sound itself because the Swarmandal, because of the way that the sound resonates in the body of that mahogany body, I was not able to just simply sample it.

Charlie Chapman:

There's two things we're talking about here. One is all the AI neural network stuff about the creativity of what to play. But the other side, which to me sounds just as ambitious, unless there's like frameworks or something which maybe is what you're using, is synthesizing the sound itself.

Sandeep Ranade:

Yes. I was only able to get so far with sampling it.

Charlie Chapman:

Were you recording your own samples, or were these stock elements or something that you had found?

Sandeep Ranade:

No. So, I have a Swarmandal here, so I recorded individual strings and...

Charlie Chapman:

You said to pull the whole thing yourself. My goodness. Okay.

Sandeep Ranade:

I mean, it was easy for me to do and play around with it and see if it works. I was hoping that it would work, but it lacked something. It lacked the characteristic of that rich, resonance sound. So, I realized that just sampling is not sufficient. I have to simulate, I have to model the interactions of these strings with the body of the thing and the room, and suddenly it became a much more full-blown problem to solve. And so, I jumped into that. I had tried to solve this problem in 2012 on Android because I was at Google, I had an Android phone, I learned Java for the sake of it.

I didn't enjoy that, but I realized very quickly that the audio is stuttering so badly because when you're playing a Swarmandal, these strings are playing, you're going to strum the whole thing, 40 strings or 30 strings in half a second or a quarter of a second. So, these are 10, 20, 30 milliseconds apart, these strings. And it was stuttering so badly that it sounded terrible. It sounded horrendous. I went to buildings across to the Android building and I had a few friends there and I said, "This is the code I have. This is the output. This is what I expect it to sound like. Is it my bug?" Because I did a lot of binary debugging to figure out where the issue is. And I couldn't detect it in the code itself.

So, it had to be like a kernel level or a OS level thing. And they said, "Yeah, Android is not designed for real-time audio at all." And I measured it and there is an audio latency from the point that I say play a sample to the point it actually hits the sound card and plays the sample. There is a 20 to 200 millisecond latency, which is not predictable. This was a deal breaker because I wanted something that is sub-millisecond because if the strings are playing of the samples are playing at few millisecond resolution, having an error of 20 milliseconds Which is the best device was not possible.

Charlie Chapman:

Android doesn't have a real time audio API?

Sandeep Ranade:

No.

Charlie Chapman:

How do games do that? Because games have to be that, right?

Sandeep Ranade:

Because they loop, they buffer everything and it's not auto generated. It's not spontaneous. So, what they do is they have short loops and as the scene changes or as the evil, whatever character comes in, the music changes. Those, it's okay to have a one time penalty of 20, 200 milliseconds. Or they have all of this playing at the same time already buffered and they just mute tracks. I couldn't do any of that because all of my music that the app generates was all based on the analysis that had happened. And so, the analysis had to be very fast on so many different parameters.

And then it had to play all of these samples correctly and at the right time because if it was delayed, either it would stutter or it would be too late. So, I got my first iPhone and MacBook when my wife quit Apple and we moved back to India. I got a nice discount on that. Actually, the tuner was a different... I didn't associate it to low latency, but it was that. But then I realized that I didn't have high expectations initially because I thought if Android has this fatal flaw, why would Apple be any better? But I was so wrong and it was so stunning even with simple code. Just playing the samples, it sounded so fantastic, so smooth, so lifelike.

And then when I modeled it, there was more computation, there was more intense stuff happening. It still sounded fantastic. The latency was sub millisecond. And now I kept adding instruments. I kept pushing the boundaries and it still sounds phenomenal and there is no other platform that can support the things that I'm doing. And that is the reason that the app is iOS exclusive, only iPhones and iPads.

Charlie Chapman:

Okay. So, at this stage, you had an app in the store that was making some money, but you were still thinking of it as a side.

Sandeep Ranade:

It was barely anything. So, I never considered it as a source of income at all.

Charlie Chapman:

So, then what you were doing at this stage, building all of this AI performer basically, were you just thinking of it in terms of for you doing performances like, "Oh, this will be really handy?" Or were you thinking like, "Oh, this will be fun to add into that app?"

Sandeep Ranade:

No, so I think that goes down to the philosophy of my problem solving. And I typically tend to solve problems for me or people that are close to me like my student. And then if it works well, then I think maybe this is useful to other people as well. But because I am the first customer of my solution, that I think for me, at least it makes it better because then you have a very high bar, you have a very high expectation of the solution and there are no compromises. So, it has to behave a certain way. It has to do a certain task in a certain way and nothing less than that will be acceptable. And so, I had a Swarmandal now, my first instrument, AI instrument.

Then I added the drones, the tanpuras, which also had the same problem because most of these drones either were synthesized or had samples, but never really took the modeling equation into account. And that was also a fun exercise which involved deriving partial differential equations of how the string is hitting the curve bridge of the tanpura, which is what makes it special that buzzing sound, the tanpura characteristic sound is because it's a curve...

Charlie Chapman:

What does that mean? I guess I'm not fully following what you even mean.

Sandeep Ranade:

So, part of this is secret sauce in some sense because-

Charlie Chapman:

That's fair.

Sandeep Ranade:

... I will extract the essence of the samples and then take that as my genetic material to then model the interactions of these sounds. And some of them are partial differential equations. Some of them are more interesting modeling based on natural phenomenon. Sometimes it is the model of the gas states bouncing around. Sometimes that works, sometimes it is elasticity that works, sometimes it is other spring properties that work. I don't know. I just try these things out. Some of them stick, some of them don't. It's not any conventional way of looking at it. It's all homegrown new ideas.

Charlie Chapman:

This is all Swift running on the CPU or are you doing stuff on the GPU at this point?

Sandeep Ranade:

So, I do use Accelerate for the really high performance parts. I will use Accelerate to run things on the GPU. There is some metal tricks that I play once in a while and there is the neural engine that I will also use and then the CPU. So, it's a mix of things. It's really impressive the hardware capabilities that Apple has created that allows me to throw workloads at it and it doesn't even break into a sweat really. It is really impressive.

Charlie Chapman:

Okay. So, you started adding instruments, multiple instruments at this point. Did you release them like instrument by instrument and you're just funneling this into the app? Or was this this big side project that you were working on that eventually you added into the app as a big new release?

Sandeep Ranade:

No, no, because the tuner was part of the app and then the tuner was integral to all of these other features. So, then the first feature was the tuner. The second feature was the Swarmandal, third version added tanpuras and 100 Indian ragas and some Western scales. And so, it progressed from there. And I now realize that there is a very ambitious project that I want to do, which is to emulate the tabla. And tabla is a very complex instrument, very capable instrument. It has a very wide variety of sounds that it can produce. And there are a lot of rules of accompaniment, both written and unwritten, of aesthetics, of appropriateness, and it's an entire PhD topic in itself.

It's made of two drum pieces and it's a vertical standing drum and you play it with your hands and the range of sounds that it can create is just absolutely breathtaking. So, I didn't know if it'll ever work or not, but I wanted to give it a shot because when I'm on stage performing live with my tabla player friends and with my harmonium friends and all of the instruments that I will perform with, that is truly magical, that is that spontaneous creation, that shared musical dialogue is so amazing, so much fun, and it elevates everyone's music to the next level. It is the best example of collaboration really that happens in front of your eyes and ears.

That magical element is solely missing when I practice by myself at home. Yeah. Because with all of these loop-based accompaniment, it doesn't actually change and it doesn't create that spontaneous give and take.

Charlie Chapman:

Every time you add an instrument then, are the AI players, are they listening to each other and doing what you just described, like collaborating, and then they're listening to your vocals and all that's going into this algorithmic soup to cause that pseudo collaboration?

Sandeep Ranade:

That is a very perceptive question because that's exactly what it does. They are listening to each other. They're loosely coupled. So, they are listening to each other, they're listening to the singer or the instrumentalist, the main artist, the user of the app, and they are creating music that spontaneously creates order. It emerges. It's an immersion property.

Charlie Chapman:

This isn't a single composition where the system is spitting out music for each of these players. You have three separate entities doing their own thing and taking each other as inputs. Exactly.

Sandeep Ranade:

And they have guiding principles. So, you set the scale or the rag, you set the tempo, you set the rhythmic cycle, whether it's a particular tal or groove, whatever it is, right? Those parameters are defined and those are given to the app. But between that, they'll basically figure out what to play, how to improvise, how to interact, when to play what, when not to play what, all loosely coupled and spontaneously. Because I thought that maintaining that artistic spontaneity would be the crux of recreating that artistic spontaneous experience that I have with my musician bodies at home.

And I think I was right at the end of it because I've never tried anything else, but this worked so well that I didn't have to try anything else. And so, the tabla became my most challenging task to date. And it took me several years because I never learned tabla. To play tabla is a whole other ball game. As a singer, I understand tabla well enough that I can sing with it, but to play it is a completely different thing. I understood enough to implement it. I still cannot play. I will never be able to play the tabla well, but I understand the nuances much better now and the app understands them much better now. And it started to grow. And here's the thing, right?

All of the new AI models, the generative AI, the LLMs, these are based on a very large amount of data. That data piece is so critical and these are huge models, gigabytes, terabytes of data is fed into it. I didn't have those resources to begin with. I didn't want to go that route because it becomes challenging legally as well, ethically as well. So, I didn't want to enter that area at all. So, I wanted to approach it in a way where when I teach my students how they learn, how the brain actually learns. So, I created something which I call the mimicist engine, which is a new way of looking at how these learnings are incorporated.

If you think of neural networks and the different weights that are associated with these neural pathways, the mimicist engine does things in a different way. This is all secret sauce. It's just a name that I want to throw out there because it is a completely new way of doing things. And then instead of neural networks, I use something called, that I call biofeedback neural networks. So, they work in a different way. And I was able to build it on top of the neural engine with Apple's hardware. And so, this is why I call it artistic intelligence.

Charlie Chapman:

Artistic intelligence. I like it.

Sandeep Ranade:

And so, that was the tabla. And then I added in more harmony instruments with the piano. There was violins, string section, trumpet section. Then I added some percussion, which was not as hard of a problem certainly. But then with 10 instruments, I had to add a orchestrator layers, a conductor level AI to make sure that all of these pieces, all of these musicians are interacting seamlessly and collaborating well without stepping on anybody else's toes or anybody else's music.

Charlie Chapman:

What was happening before you had that? Were you starting to notice they were playing over each other or they couldn't...

Sandeep Ranade:

Because it was loosely coupled before and it is still mostly coupled, but now it has a little bit of a guidance. Suddenly the music would swell in, suddenly it would diminish because everybody would start playing at once and then everybody would drop off. Not everybody, but like the string section would come in with the trumpet section and create a peak and then it would suddenly go down. And that didn't sound great. So, now they collaborate within each group as well as with the outer, with the rhythm and with the singer and with the other things. So, there are ways of composing in layers with these instruments.

Charlie Chapman:

Okay. So, you just described you've added a whole bunch of new instruments, you added this conductor layer. At this point in the story, is it still just an app that's in the store that's not really making any real money or has it started gaining traction and getting attention by this point?

Sandeep Ranade:

So, by this point, actually with the 10 instruments, this was around the time that the pandemic hit. And during the pandemic, I was feeling very helpless and depressed and because I wanted to help, but what can I do? I'm a software engineer, I'm a singer. How can I possibly help in this dark time? And so, then I thought maybe I can create a song that encodes tips of how to survive the pandemic and then also put in some optimism that yes, we will survive the pandemic. Although we didn't know it at that point in time, but it was such a depressing time for everybody that I knew that I thought some optimism would be much needed.

And so, I did this song, I composed it and I recorded it with the app. And this was a messy setup because it took me about two, two and a half hours to record a 2-minute song with the setup with the app putting out one single audio track and then I had to record my voice separately and then video separately and then mix it in.

Charlie Chapman:

Because you have a YouTube channel now that you do videos and stuff like this.

Sandeep Ranade:

Yeah, I had a YouTube channel for a while and I put it up there, but I also sent it out to a few friends on WhatsApp, like the video I sent it on WhatsApp. It's the first time that I'd done this. And to my utter astonishment, it went viral in a way that... I mean, you hear stories about this, but you never think that it'll happen to you. And it went viral. My phone started blowing up 10 minutes later and I got a call from my Guruji. You have to realize that Pandit Jasraj was, he's no more, but he was an icon. He was one of the most famous musicians of the world at that point in time, a career that span nearly 70, 80 years. He was nearly 90 years at that point in time and incredible musician, composer, phenomenal person.

I had not had the courage to send him my composition because it was a classical Hindustani classical song in Raag Basant, but the main topic was a virus and it had some geeky puns and some humor and I was not sure. So, I only sent it to a few close friends to see what their reaction would be, and then I decided to put it up on YouTube. Before I had actually put it up on YouTube, he called me and he said, "At least 300 people have sent me your song and I'm wondering who are these 300 courageous people who can send a video forward on WhatsApp to Pandit Jasraj." And then he said which means that you made spring blossom and it was a play on words on the rag name as well. And then he asked me the question.

tabla has been played very well. Who has played the tabla with you? Who? My app had passed the musical turning test. This was a huge milestone for me.

Charlie Chapman:

Wow. So, he was asking you who played that instrument.

Sandeep Ranade:

He didn't know about this work because it was recent work.

Charlie Chapman:

Of course. Yeah.

Sandeep Ranade:

And he didn't know about the capabilities of the app and what I was doing by that time. And then I told him and he was so happy that he said, this is very necessary work. You're doing the right thing. You're going to keep music alive and keep it evolving. And he gave me a lot of blessings, but that moment was-

Charlie Chapman:

Wow.

Sandeep Ranade:

... so much... And from a judge, like if you imagine them during test, there's a judge, there's a human judge and there's a computer. You cannot expect a better judge than Pandit Jasraj.

Charlie Chapman:

Have you run into any of the fear or negativity about you're going to replace real musicians?

Sandeep Ranade:

Absolutely. So, I wanted to answer that question about the revenue and the app going well. So, after this point, people started realizing that this was actually accompaniment by an app and then they started looking into it and I saw uptake in the revenue as well and the downloads. But yes, there were a lot of people who initially thought that I was doing something wrong by infusing music with artificial intelligence, that I would replace accompanists and artists. And this was a few years ago now that fear has been completely superseded by what other AIs are doing, which is completely replacing the entire process itself.

And you just give it a prompt and it gives you the song completely in a given style, voice, with all the orchestra. But I don't think that is fully music because it is still bits and pieces from other artists that have been merged together in perhaps new ways. But what my app was doing was enhancing humans. When these people saw what this AI is now doing, the new AI, they realize the value of my app and they realize that I'm giving them a flight simulator. I'm not giving them a self-driving car. I don't want to replace the drivers.

Because see, when you're on stage, that experience comes once in a while for the vast majority of musicians, only the most professional, the most frequent performers will get onstage experiences every day, but the vast majority will have to practice at home, hoping that they will have accompanists once in a while to bounce their musical ideas off and collaborate with, and then they'll be comfortable on stage. So, it's like a flight simulator because if you don't have access to a plane, then you want to practice in an environment that is as close, as realistic as possible, so that when you do get access to a plane, you not only survive it, but you also enjoy the ride.

And so, that was my intention, that was my thought process, and now people are seeing it. And in fact, there is a lot of ethical boundaries that I've put in the app for that very reason. Because for example, if there is a tabla phrase or a drum phrase or a guitar phrase, the app can certainly play faster or more complex than a human can ever achieve. But I've put limitations based on what a human can do so that the app still feels realistic and human.

I have to compute how fast, how far the hand will move on a surface like the tabla with the friction involved, and then how fast can certain combinations of samples or bowls can be played or on a guitar, if you're strumming, how fast that the wrist can actually move. I've mapped out great performers and put these boundaries and put these limitations into the app so that it feels natural, it feels realistic.

Charlie Chapman:

Yeah. Oh, that's crazy. So, at this stage then, it's still paid upfront, but it's starting to get some attention and some traction?

Sandeep Ranade:

No, by this time I actually changed my model. So, the tuner was pay upfront, but the Swarmandal onwards, the second version onwards, Apple had just introduced subscriptions. So, I made the app free and you can try the features out as many times as you want. Some instruments, some features will be time limited so that you get an idea of what the feature is. And if you think that it is useful, then you can subscribe. It's a small subscription. It's like a medium coffee latte per month.

Charlie Chapman:

It's a lot of professional or at least hobbyists, like people taking this very seriously as a practice, right?

Sandeep Ranade:

Oh, absolutely. And in fact, I looked at some of the analytics. The average number of hours per person per day that the people are using the app is like four hours, 50 minutes-

Charlie Chapman:

Oh, man.

Sandeep Ranade:

... in over eight sessions. So, these are serious musicians.

Charlie Chapman:

Yeah, don't let the VCs on the call hear this.

Sandeep Ranade:

I've had offers to buy out the app, and the only question I ask them every time is, "What feature would you add next?" They had no answer. And it now truly seems that I'm the only person who's passionate enough to evolve this app in the way that I care about, in the sense that an artist, a performing artist cares about. Because there are things that people don't know that I've done that will not matter to the vast measure out of people, but it'll matter to me like the tanpuras getting all of the partial differential equations, all of that. I mean, it's not truly necessary from a business point of view, but from an artistic point of view, that really makes the world of a difference to me.

Charlie Chapman:

But there has been some recognition that you've gotten because I think maybe you can see in the background of your frame over there, there's a little aluminum or aluminium cube over there. What's the story behind that?

Sandeep Ranade:

So, I think the dream started in 2019 because that was the time that I got to actually go to the WWDC Apple's worldwide developer conference in person. And after that, there was a pandemic. So, that was the first year that I was there and then it became online after that for a few years. And I attended the Apple Design Award ceremony there as an audience member. I was looking at all of these beautifully designed apps being showcased on the screen and they got this superb cube. And there was like a split second where I thought one day perhaps. And then I quickly threw that out because I realized that I'm a niche developer.

I'm a hobby developer working in a very niche area of Indian music and classical music at that point. This was before all the guitars and drums and everything else came in. And I realized that I don't really have design expertise. I'm a fan, I'm a enthusiast of design, but I'm not professionally trained. I'm not a career. I don't have a professional career in design. I don't have a marketing team. I don't have a design team. These are all superly designed, expertly designed apps that are winning. So, I assumed that this is not for me and I forgot about it, but forgot it was probably still there at the back of my mind, but I thought it was not realistic, but it really felt like I was at the Oscars.

That experience was so inspiring and overwhelming just as an audience member. So, there was probably that at the back of my mind. And I continued adding features. And in 2021, not Southerner won for innovation, the Apple Design Award, which was amazing for many reasons because I'm only the second Indian to ever win this award. And so, from that point of view, it was a big deal for me. But the more important thing was that this was the very first time that Apple was giving this award for an app that was about Indian music and culture. And so, as an Indian musician developer, it really meant the world to me that Apple was able to recognize this effort.

Charlie Chapman:

Whenever I was in Bangalore, that's where we met, was at Swift Bharat. I really got the sense from the developer community and the Apple developer community, at least, that there was a real sense of pride in the emerging Indian developer scene. Since I've been doing it, there's been a long list of Indian Indies, that's funny to say together that been making really cool apps. And at least on Twitter, I've known a whole bunch, but there was a real sense whenever I was there that there's this groundswell and it seems like it just started with you. And I guess Raj was...

Sandeep Ranade:

Raja was the first Indian to win there. 2018, I think.

Charlie Chapman:

When I think about my list of who has a chance for an ADA in the future, there's a couple Indian developer designers that are definitely clearly on that list.

Sandeep Ranade:

See, now Indian colleges, there are some four or five universities now that have iOS developer centers in the campus itself. So, they have a hundred Macs, IMAX and iPhones and trained staff, mentors, and they are developing apps a couple of decades before I started my journey and they have so many more resources now. And so, I'm sure that there are going to be some incredible apps, some amazing innovations that come out of that ecosystem, not just India, all over the world, because students are now seeing the Swift Student Challenge and all of these programs. There's so much awareness at a young age. So, it's really exciting to see what I can learn from what they are doing and what inspiration I can take from them.

Charlie Chapman:

Yeah, that's super cool. But yeah, I mean, getting an ADA is a crazy achievement. I guess that was 2021. So, that would've been during the virtual time. So, you didn't get to go on stage and...

Sandeep Ranade:

No, I got a call at 2:00 AM and say, "Are you awake? Are you watching the ceremony?" I'm like, "Yes, stay awake. Stay watching the ceremony." And then lo and behold, the names were announced and I was there and it was surreal. I sometimes still can't believe it. It's sunken a little bit now, but it's been a few years, but it was completely surreal.

Charlie Chapman:

Ah, that's so cool. So, I guess coming off of that, did that make a dent in the business? Did it feel like things started picking up even more after that?

Sandeep Ranade:

Yes. So, some of them were casual downloads who see just out of curiosity what an Apple design about winning app looks like and perhaps... So, some of them didn't...

Charlie Chapman:

Yeah, your conversion rate probably dropped quite a bit. Yeah.

Sandeep Ranade:

Yeah. So, there was a rather and then a small, say 20%, maybe 15% dip. But what happened was a lot of younger generation who were tuned into the Apple Design Awards and knew about it. They downloaded it. Some of them were musicians or students of music and they told their teachers that there is this new app that has come out and we should try it out. And there were a lot of artists, truly phenomenal artists who looked at this app, who played with it, who understood the value that it brought to the musician community as a whole, and they recommended it to their students. And so, that is how it started to then permeate the musician, musical ecosystems.

Charlie Chapman:

So, at this point though, were you full-time on this as your main job or were you still doing other things before that or at that point?

Sandeep Ranade:

Oh, I was doing several full-time jobs. So, I have my two twins at that point, and this was a full-time job because the app started two days after they were born. So, I was developing either when they were sleeping or I was practicing the Swarmandal stuff and music and to put them to sleep. And so, wherever I got few minutes, five minutes, I would add a line of code, I would add a small feature, I would add a test. And it has been like that. It's been getting better as they're growing older, so now they're almost eight. I also am a performing artist, so I travel and perform at different parts of the world. I've been giving talks on my journey on... I become a motivational speaker somehow.

So, that has been also occupying my time. It's been increasing in frequency. And then the endless evolution journey was also continuing. There were a couple of corporate gigs that I would do once in a while. And so, it was a mix of all of these things, but I wanted to spend the large part of the day with my kids. The primary reason I left my job and my wife left her job, and people actually were not happy with our decision then. A lot of people said, "Oh, being really foolish to leave a lucrative job at Google and Apple in the US and then moving back to India." But we wanted to be here for our kids as much as we can. And so, that there are no regrets.

Absolutely, it's been a wonderful thing, but wherever I get the time I will code, I will do. There's also content creation. So, I'm recording, I'm creating original compositions and content for YouTube and other channels. So, there's all of that. And I'm also teaching.

Charlie Chapman:

At this point though, this app and the stuff around it is your main job then outside of family, obviously.

Sandeep Ranade:

Yeah, effectively because all of it feeds into that. My teaching, my music, my creation, all of that feed into the app now. So, it becomes this one-stop-shop kind of app for all of my proclivities.

Charlie Chapman:

That's fun. So, as this became your main source of income then, were you doing things to grow the business? Were you thinking of it in terms of a business or were you pretty much focusing on the product and then that people were coming and so you could just keep focusing on the product and that drove the growth of the business?

Sandeep Ranade:

I think I was focused more on the feature set and capability of the app because in my mind, and this is perhaps a mistake or a naive way of looking at it Right? But I thought that the more capable I make the app, the more attractive it'll be to people to use and to use on a daily basis, and then they might subscribe to the app. That has been happening. The app has been growing steadily and I've not done any targeted marketing. There might be a time that I do targeted marketing, but I keep thinking that there is this feature to add, there is that feature to add.

I don't want to make the mistake of marketing too early, but now I think there is sufficient amount of features in the app that perhaps I started thinking about it a little bit. So, maybe in the future I can do that. But yeah, so far it's all been organic word of mouth.

Charlie Chapman:

That's what I was going to say. Where's it coming from? It's not like ASO search traffic. No. It makes sense here because people, you've invented a category. So, if you're not doing the marketing yourself to educate people to come here, then they have to be coming from, I guess, yeah, word of mouth or users, which I guess maybe this is word of mouth, but users sharing it themselves in public channels?

Sandeep Ranade:

Yes, exactly. Because you have to understand how good it is for your music to be able to tell other people that it'll help them with any credibility. And the other thing is that if I create videos, the accompaniment is now so lifelike. The music that it creates is so lifelike and so good that you might not even believe that it is actually really done by an app. You could think that it was post-processing or somebody played a track for you or whatever.

Charlie Chapman:

That's the thing, right? As a demo, it sounds like you're just playing prerecorded music.

Sandeep Ranade:

That's the problem. It's so good now that it feels like it's not really true, but only when I do live demonstrations, people really understand that, oh, this is all real time, this is all spontaneous. This is all happening on the device locally. And then they really get so excited. So, the live demo that I did when we did the live version of the show, that audience was so excited about it. They were so in the moment in the music.

Charlie Chapman:

So, at the live show we did, we ended it with you demoing the app live on stage. It was very exciting. It was an Indian conference, but there was a decent international crowd. This is actually my question where it had Harry Potter thematic elements to it, but it was obviously

Sandeep Ranade:

It was the Harry Potter theme of Hedwig composed by the great John Williams, but I had added my own Indian fusion twist to it. So, when I was singing, it was the Harry Potter theme. But what I tell the app is what is the rhythm, what is the tempo and what is the scale? And so, there's a rag called Kirwani, which fits very nicely within that part of the theme. So, I told it that. And based on that, it was able to listen to me and then build everything else spontaneously. So, if I do it again, it'll sound different. It'll sound like the Harry Potter, it'll sound appropriate, but the elements will be different each time.

Charlie Chapman:

Okay. So, it was your singing that brought in... What did you call it?

Sandeep Ranade:

The rag. So, it's like a scale. It's a set of notes, but it is much more than a scale. There are a lot of rules involved. There are some nuances, there are some traffic rules. It makes it a state machine of notes that has proportions, that has probabilities, that has some predator mind parts, some spontaneous parts. It's a complex state machine. On the other end, a scale is usually very straightforward. It's a fully connected graph and it's a state machine where any state can follow any other state legally. In a rag, that might not be the case always.

Charlie Chapman:

So, you were just selecting a rag that Hedwig's theme fits in well with, and then you were singing, obviously knowing the theme, and then the rest of it was just the band, the AI band keeping up with you.

Sandeep Ranade:

Yeah. It was synthesizing all of that music on multiple instruments on the fly.

Charlie Chapman:

Wow.

Sandeep Ranade:

I did spontaneous elements at the end of it. I did some improvisations and I said, sang some sargam, some solfege. I was just in the moment and it followed me, accompanied me precisely well through all of these unplanned moments. And in fact, I did that to highlight that this is not a karaoke track or this is not prerecorded. This is actually live.

Charlie Chapman:

We talked about this ahead of time, and I think maybe we could try this now. Let's see if we can do a live demo right here. Are you still down for that?

Sandeep Ranade:

Yeah.

Charlie Chapman:

Okay. So, you're going to pull out your phone and hopefully we will record the track through the phone and then you'll be singing it live and we'll put them together and we'll get a little demo of what we're talking about here.

Sandeep Ranade:

Sounds great.

Charlie Chapman:

And if you're just listening, you're not watching the video. Pretty much, it's going to be you holding your phone. And once you start it's not like you're sitting there hitting instruments or anything. It's just going and it's using the microphone to listen to what you're singing and keep up, right?

Sandeep Ranade:

Exactly. Once I start the instruments, everything that they play till I tell them to stop is all spontaneously decided by the instruments based on what it hears me do.

Charlie Chapman:

All right. Let's hear it.

Sandeep Ranade:

I started the melody harmony. I'm going to start the rhythm. (singing) Now here is the improvised part. (singing)

Charlie Chapman:

Oh, man. When you did that in that theater, it was crazy. Everybody was going nuts. That was so cool. We've pretty much covered the story. Is there anything else in particular that you want to make sure we hit on before I start the ending bits here?

Sandeep Ranade:

I think one of the things that was very traumatic to some people that I knew at Apple was after they announced the Apple Design Award, I was thinking of rewriting the app completely from ground up and they basically said, "Don't be foolish because who redesigns an Apple design about winning app?" And I said, "I have to do this because that is the only way that I see it go to the next level." Because by this time there were 10 instruments, but to add the next set of rhythm instruments or percussion instruments, like guitars or drums or another tabla or a cajón, like a clap box, the architecture didn't support it. So, I decided to, I threw out everything I had done for those five and a half years, six years.

Charlie Chapman:

On the UI side?

Sandeep Ranade:

No, everything.

Charlie Chapman:

Oh, wow. Okay.

Sandeep Ranade:

Everything. I basically took that project and all the files and I put it in archive and I started a fresh project in Excore from scratch and it had taken me six years to get to 10 instruments. And when I restarted everything, it took me nine months to go rebuild not only all of those features again, but now reach 27 instruments and perhaps a hundred new features. So, there were workouts, there was songs and playlists like presets, there was a keyboard, there was karaoke stuff, like you can change the pitch and the speed and all of that. And then I also added the recording studio in. So, this is a fully multi-track studio quality audio, HD video, recording studio, and it also does automatic mixing.

So, really as a creator, it becomes my best friend because I told you that recording for that Na Corona Karo song, the Coronavirus Bandesh, the song took me nearly two and a half hours to do a 2-minute song and I'm very lazy. I'm like, "This should not take so much time."

Charlie Chapman:

Says the guy who's like, "I'm going to rebuild my entire app and spend nine months..."

Sandeep Ranade:

No, because that was the laziest way to get to the next part. There was no refactor possible. There was no incremental change possible that would get me there." So, it was lazy. I don't like to repeat effort. I'm willing to spend the time and study and do it well once, and then it should automate. So, I really wanted to automate my mixing because if you've mixed any songs, you know that as the number of tracks, the number of instruments increase, it becomes exponentially harder to mix well. And it's almost like black magic, like a dark art. It's an art more than a science sometimes. I feel I got good at mixing in Logic Pro and so on, but it still took me like a day to mix 27 instrument tracks.

And so, I thought maybe this is something that the app can do better than I can do. And so, it now builds a profile of the mix as it is a recording and then applies it and it is a dynamic mix. So, as the instruments come and go, the mix will automatically adapt and create the best possible sound in those whatever tracks are playing at that point in time. So, it saves me a lot of time. And now content creators are discovering this and creating phenomenal recordings with different combinations of these 27 instruments and it is just a joy to behold them. Behold, them creating something so spectacular and it's 6-year olds to 90 year olds. It is Indian musicians, Western musicians.

There are Carnatic classical musicians, instrumentalists, dancers, even, composers. There are people who don't have vision, who are the app is now fully inclusive, fully accessible with voiceover for colorblind people as well, for people with partial hearing loss.

Charlie Chapman:

Accessibility is an area that you're super passionate about, right?

Sandeep Ranade:

Yes. I think the vision is music without any barriers, without any boundaries. And so, just because a person doesn't have eyesight shouldn't make it a boundary, shouldn't make it a barrier to be a musician or to create content.

Charlie Chapman:

You built all these tools that make it easier to create content, partially it sounds like for yourself, but I have to imagine that's feeding into that loop of content creators creating things with the app helps spread the word about the app.

Sandeep Ranade:

Absolutely.

Charlie Chapman:

Which then gets more people to know about it, which then make content and it self-perpetuates to a degree. But before I let you go, I have to ask you the question that I ask everyone to end the show, which is, what's a person or people out there that inspire you in your work that you'd recommend others check out?

Sandeep Ranade:

There are so many sources of inspiration, but since this is more of a iOS, Apple a flavor, there's of course the point free dudes, Brandon and Stefan, I enjoy there. I think I'm attracted to people with very interesting philosophies about problem solving. So, right from the extra all the way to these new people who have a strong opinion about something, about problem solving.

Charlie Chapman:

And systematize that thinking, right?

Sandeep Ranade:

Yes. So, their work, I deeply appreciate, but I think there's one person who made an impact on my way of looking at problem solving, and that is Bret Victor. ex-Apple employee, he has done a bunch of design work, he's an engineer, but he has a couple of very powerful videos on YouTube. There's the inventing on principle, there is I think the future of computing or computation, and there is a couple of others. But I think what he always talked about that really resonated with me was that creators need to have this instant connection to their creation.

So, if you are building cohort and then you have to compile it and then you have to run it and then you have to create the test environment and then test it to know whether what you have done has worked or not. That process has a lot of buffer in the middle as opposed to an animator or a musician changing a knob on something and immediately knowing that yes, that instrument volume changed or the reverb amount changed or things like that. So, I think SwiftUI, playgrounds, previews, all of these were probably inspired from his ideology, his work. And then for me also when I create UI, to see what works, there are so many parameters, there are colors and saturations and shapes and sizes and corner radiuses and whatnot.

From that point of view, if you have a set of tweaks, sliders that you can quickly change as opposed to changing code and you know what looks good and then the tools will basically generate the code for you and put it in. Or from an AI point of view, something like parametric control, gesture-based controls, physical modeling of instruments, the simulation of different parameters for different AI models. If all of these were instantly actionable, the feedback is instant, then it makes the creation process so much faster, so much better. So, I've put all of this in practice in my app and it really saves a lot of time and effort. The iteration process becomes so much more rapid.

So, I think his video is certainly your listeners should check out and it's a very powerful set of ideas.

Charlie Chapman:

Yeah. I'll definitely include these in the show notes. I don't think I've come across these before, but it looks like they're older. I see 2012 was the Inventing on Principle talk.

Sandeep Ranade:

Yeah.

Charlie Chapman:

So, yeah, they've been around for a while. So, this definitely is going on my list of homework.

Sandeep Ranade:

I came across these, I think a couple of years into my app development process. I don't know how I came across it, but some rabbit hole.

Charlie Chapman:

All right. Well, it was seriously super fun to get to hang out one-on-one, not in front of 300 of our other closest Indian friends. So, thank you. Thank you very much for coming on.

Sandeep Ranade:

Thank you. It was a wonderful conversation and you pulled out a lot of things that I've not really had the chance to share with a lot of people before because I didn't really think that they would be interested in a deep dive in psychoacoustics, for example.

Charlie Chapman:

The question I always have with this show is, is anybody else interested? I don't know. All I know is that I get very interested, so I will pull on those threads.

Sandeep Ranade:

Wonderful to see Charlie unplugged. And I had a wonderful time. Thank you again for having me on your show.

Charlie Chapman:

Awesome. Well, where can people find you and your work?

Sandeep Ranade:

So, my website I think has all the links to all of the different stuff that I'm working on. It's Sandeepranade.com and that's it.

Charlie Chapman:

Sweet. We'll have that in the show notes.

Sandeep Ranade:

Wonderful.

Charlie Chapman:

Thank you again, Sandeep, and I will catch you hopefully at some other point somewhere in the world.

Sandeep Ranade:

I look forward to that.

Charlie Chapman:

Thank you so much for listening. You can find more Launched at launchedfm.com, and you can find me on pretty much all the social medias. I'm @_chuckyc on Twitter or Charlie in Chapman pretty much everywhere else. And of course, huge thanks to RevenueCat for making this episode and all future episodes of Launched possible. I'll see you all again in two weeks.