S4E16: Carlos Mbendera

Summary

Slices: The Deep Dish Swift Podcast

Episode S4E16: Carlos Bandera (April 9, 2026)

---

Episode Overview

In this episode, hosts Kai Dombrowski and Malin Sundberg sit down with Carlos Bandera, a former Apple engineering intern, WWDC Swift Student Challenge winner, and indie developer. Carlos shares his global upbringing, his journey into tech and iOS development, the creation of his music discovery app Setu, and insights into the rapidly evolving landscape of audio APIs and AI in software engineering. The conversation covers career choices, indie app development strategies, AI’s impact on university studies, and Carlos’s upcoming talk at Deep Dish Swift 2026.

---

Key Discussion Points & Insights

Carlos's International Background and Journey to Kansas

• Early Life & Global Experience
  • Carlos was born in Malawi and raised across South Africa, the UK, China, and the US.
    • “I grew up in like Malawi, South Africa, the uk, China, and here.” (00:42)
  • Moved frequently due to his father’s work as a consultant.
  • Recently spent a semester abroad at Korea University and traveled to Japan.
    • “I did a study abroad this past winter. I was at Korean University … went to Japan later afterwards.” (01:03)
• Adapting to New Cultures
  • Carlos reflects that frequent relocations were easier as a child because making friends was his main focus.
  • Experienced the COVID-19 pandemic while studying in China (12th grade), which brought unique challenges.
• Why Kansas?
  • Chose University of Kansas for a generous scholarship and appreciation for the Midwest.
    • “Kansas? Well, I got a really good scholarship from my university, and… it's just been a fun time.” (02:30)

Route into Tech and iOS Development

• Introduction to Programming
  • Inspired by the TV show Silicon Valley to enter tech:
    • “I was like 15 or 14… I was like, cool, so you can make an app and it can get, like, super famous and popular. I can do this too.” (05:01)
  • Started with a $10 Java course on Coursera, building web apps before moving into Swift.
• WWDC Swift Student Challenge
  • Began iOS development to solve a personal problem: wanted a better manga reader for iPad.
  • Discovered Paul Hudson’s “100 Days of SwiftUI.”
  • Entered and won the WWDC Swift Student Challenge with a rhythm trainer app using computer vision.
    • “It won the first year. And afterwards… [I] tried in this Apple internship stuff. And then that happened to work out…” (04:08)

Internships and Apple Experience

• Apple Internships
  • Applied for Apple internship after Student Challenge; was surprised to receive an offer and completed two internships at Cupertino.
  • Notes how different universities require (or don’t require) internships for graduation.
  • “For me personally, I kind of just gave it a go, and I was like, let me submit my resume. What's the worst that can happen? And it just happened to work out.” (03:52)

Indie App Development: Reze & Setu

• First App: Reze — AI Job Assistant
  • On-device Mac app using Apple’s foundation models; analyzed resumes and job descriptions.
    • “It did okay. It wasn't a great app, but it was a fun project.” (06:27)
• Second App: Setu — AI DJ & Music Discovery
  • Originated from Stanford’s Treehacks Hackathon, initially started as an AI DJ concept.
  • Iterated with user feedback (primarily from Reddit and Apple Music community), quickly reaching over 1,000 downloads after launch.
  • Setu’s audience: Apple Music users who care about intentional listening and want better music discovery than Apple’s built-in algorithms offer.
    • “The main market seems to be people that really enjoy listening to music and are very intentional of their listening, but have a hard time discovering new music.” (08:49)
• Launch & Promotion
  • Used Reddit as the primary launch channel, with additional word spread via Discord, Slack, and friends.
  • For version 1.2: more proactive press outreach; submitted to App Store nominations and major Apple news outlets.
  • Monetization: Not currently monetized; Carlos treats Setu as an experiment to test the indie app pipeline before building a more scalable/commercial product in the future.

Radio Hosting and Music Discovery

• Hosted radio shows at KJHK 97 FM (Kansas), Leeds Student Radio (UK), and BayFIFTY FM (San Francisco).
  • “I used to DJ here in Kansas… and also in Europe at the LEED student radio… in San Francisco at Bayfift fm.” (12:32)
• Shows focus on international music, often curated by researching culture and musical trends (e.g., Japanese math rock).
• Relates his app Setu to his need to discover international music, previously using Apple Music, Spotify, and YouTube’s recommendation algorithms.

AI & Academic Life

• Carlos has witnessed the shift from AI as a curiosity (2022) to a core developer tool during his time in university.
  • “I've kind of had seen the entire arc of like AI being like, oh, it's this cool thing people use for their homework once once in a while. … Got to be AI first software engineer.” (16:23)
• University policies on AI use vary widely by professor; tech interns see more encouragement of AI tooling in industry.
• Noted increasing focus on AI in job application processes (e.g., Meta including comprehensive AI usage questionnaires).
  • “Recently had a reach out for Meta and they had a lot of AI questions in the email itself before even interview.” (18:39)
• Student attitudes toward AI tooling run the gamut from casual to hardcore adopters; Carlos positions himself closer to the “hardcore” end, but still selective.

Deep Dish Swift Talk: "Extra Saucy Swift Audio APIs"

• Talk Theme:
  • Introduction to Apple’s new Speech Analyzer framework (replacement for 2016’s Speech Recognizer), as well as open source alternatives.
  • “I'm giving a talk on extra Saucy Swift audio APIs. So last year Apple released a new framework called Speech Analyzer…” (22:09)
• API Improvements:
  • Speech Analyzer is modular, well-documented, and easy to use for transcription and audio feature extraction (e.g., pitch).
  • “With the new Speech Analyzer, you write like five lines of Swift code and it's done and you have like output…” (24:09)
• Incorporates open source models (e.g., for speaker diarization) and describes their modular pipeline integration.
• Emphasis on how Apple’s internal use (voice memos, notes, on-device summarization) utilizes the same API available to third-party devs.
• Plans to publish a companion article and starter project for the community.

---

Timestamps for Important Segments

• Carlos’s International Story & Adapting — 00:42 – 02:17
• Programming Origin & Silicon Valley Inspiration — 04:52 – 05:30
• Apple Internships & Student Challenge — 02:55 – 04:42
• First Published App (Reze) — 06:12 – 06:51
• Setu: Creation, Launch, & Audience — 07:20 – 09:03
• App Growth & Monetization Thoughts — 11:18 – 20:38
• Radio DJ Experience & Music Discovery — 12:22 – 15:37
• AI’s Impact on Academia & Industry Interviews — 16:23 – 19:05
• State of Student AI Adoption — 21:01 – 22:01
• Deep Dish Talk Preview: Apple Speech Analyzer — 22:09 – 25:39
• API Modularity & Integration — 25:39 – 27:53
• Inspirational Figure: Nujabes — 28:39 – 29:29
• Combining Music & Tech — 30:09 – 31:21

---

Notable Quotes & Memorable Moments

• “I was like 15 or 14... I'm like, cool, so you can make an app and it can get, like, super famous and popular. I can do this too.”
  — Carlos Bandera (05:01)

• “The main market seems to be people that really enjoy listening to music and are very intention of their listening, but have a hard time discovering new music.”
  — Carlos Bandera on Setu’s audience (08:49)

• “I feel like Apple music is very much like quality focused... But discovering new music is kind of hard.”
  — Carlos Bandera (09:17)

• “I've kind of had seen the entire arc of like AI being like, oh, it's this cool thing people use for their homework once once in a while. … Got to be AI first software engineer.”
  — Carlos Bandera on the changing AI landscape (16:23)

• “With the new Speech Analyzer, you write like five lines of Swift code and it's done and you have like output, you have transcriptions ready to go. … Five lines versus having like a view controller that's also doing some pretty strange stuff.”
  — Carlos Bandera (24:09)

• “I guess I'd have to say Najabis. Najabis was a hip hop producer from Japan in the early 2000s… It was like very cross cultural… That’s the attitude I want as I go through industry.”
  — Carlos Bandera on his biggest inspiration (28:39)

• “[Tech] is a really good kind of binding box to whatever other interests you have, which then makes you be able to uniquely build something in that space.”
  — Host 2, on the value of combining tech with other passions (30:40)

---

Episode Flow & Tone

The conversation is congenial, curious, and collaborative, blending technical details with Carlos’s personal narrative. The hosts strike a balance between inquisitive and supportive, encouraging Carlos to share both the practical details of his projects and his personal reflections on learning, ambition, and inspiration. References to international experiences, tech memes, and inside jokes (e.g., the “layup” leading to Setu) keep the tone lively.

---

Summary for Non-Listeners

In this episode, listeners meet Carlos Bandera—a global nomad, developer, and DJ—who discusses his journey from building simple apps to shipping polished, AI-powered indie apps like Setu. Carlos offers an in-depth look at navigating Apple internships, leveraging AI in creative and academic work, and the new possibilities unlocked by Apple’s Speech Analyzer APIs. His story is a testament to mixing tech skills with creative pursuits and being willing to experiment—even (and especially) when the future is uncertain. The episode closes with practical wisdom about studying tech, the importance of integrating multiple passions, and the staying power of curiosity.

Transcript

[Kai]

Hello and welcome to Slices the Deep Dish Swift podcast. On this podcast, Kai and I chat with the speakers of the Deep Dish Swift conference.

[Host 2]

Today we're talking to Carlos Bandera. Carlos is a former Apple engineering intern, WWDC Swift Student Challenge winner, and an indie developer passionate about audio, metal and combining Swift with music. Hey, Carlos, nice to meet you. This will be fun.

[Carlos Bandera]

Hey, nice to meet you too.

[Kai]

Yeah, thank you so much for being on the show and I'm looking forward to getting into all all the stuff that you've been up to. We were saying before the recording, your website looks amazing and we want to talk about the stuff you've been building. But to start off, let's just start with where you are and where you're from.

[Host 2]

Right.

[Carlos Bandera]

So I'm currently in Lawrence, Kansas. I'm at the University of Kansas. Originally, I was born in Malawi, but also grew up in like five different countries. So I grew up in like Malawi, South Africa, the uk, China, and here.

[Host 2]

What brought you to all of those places?

[Carlos Bandera]

My dad's a consultant, so he moved along.

[Kai]

Cool. I saw. Did you also study in Korea for a bit?

[Carlos Bandera]

Yeah, I did a study abroad this past winter. I was at Korean University. I got to meet some cool developers who are part of the iOS community and Apple Developer Academy that side. It was super fun to Japan later afterwards, which is a cool experience as well.

[Host 2]

How is it to move between countries that much when you're younger? Was it hard to get used to all those different places, cultures, languages?

[Carlos Bandera]

I don't know. I feel like when you're a kid, it's just like as soon as you have friends, you're like, just happy with it.

[Host 2]

Yeah. Did you learn languages of all the places like when we moved to China, that probably would have had. Did you go to an English international school there?

[Carlos Bandera]

Right. So when I was in China, it was like my 12th grade and basically did a foundation year and my classes were in English, but also took Chinese classes and also during the pandemic. So I sort of had like a very unique experience that one year in China.

[Host 2]

Wow.

[Carlos Bandera]

That is.

[Host 2]

That was an intense year in China too, because China was super strict about COVID right?

[Carlos Bandera]

Yeah, was super weird. I was like, am I going to go back to in person classes? Am I just going to be in my room forever?

[Kai]

Now, out of all the places you've been to so far, which one is your favorite?

[Carlos Bandera]

That's a very hard question. Very hard. But right now, if you would decide

[Kai]

to spend an additional year in any of those places, which one would you pick.

[Carlos Bandera]

I'd probably go with Korea because I feel like it was the most different and the most unique and the people are super friendly. The food was great, the subway was clean. It was great.

[Host 2]

And what ended up bringing you to Kansas now?

[Carlos Bandera]

Kansas? Well, I got a really good scholarship from my university, and I do like the Midwest, and it's just been a fun time, and I guess I'm kind of stuck here now.

[Host 2]

Are you studying computer science at the moment?

[Carlos Bandera]

Yeah, I'm doing computer science and I graduated in May, but I've been doing some other stuff in between, such as conference talks in the app development and also the stuff at Apple, so. Yeah, so.

[Kai]

So the Apple work that was part of internships, right?

[Carlos Bandera]

Yeah, yeah. So I started iOS development with my. The WDC Swift Student Challenge. And then afterwards I was like, oh, let me just try applying for an appointment. Maybe it'll work out. And then for some reason, a manager is like, oh, would you like to do some interviews? And then I was in Cupertino in my first summer, and I was like, then they liked me enough that I went there again a second time.

[Kai]

How does that work? Because it's different in different countries with. With internships. Does that become. Is that a requirement at your university that you should do internships, or is it sort of up to each student what they want to do?

[Carlos Bandera]

It really depends on the school. So I have some friends that go to the University of Rochester, RIT and Harvard and Waterloo, and, like, some of those schools, like, really require you to have, like, an internship. But my school is somewhat flexible. Like, you can do an internship if you want, and it comes like, your professional electives. But if you don't want to take credit for it. Some people just internships just because they want their personal work experience and personal development.

[Kai]

Yeah.

[Carlos Bandera]

Yeah. For me personally, I kind of just gave it a go, and I was like, let me submit my resume. What's the worst that can happen? And it just happened to work out.

[Host 2]

Yeah. And the Swift Student Challenge was your introduction to Apple development, or did you do stuff before that already?

[Carlos Bandera]

Right. So before university, I was really into reading manga, and I got my first iPad and I was like, I want to read some manga, but there's no good iPad app. So I'm like, let me make my own. Then I found Pohatson's 100 Days of SwiftUI, and I was like, oh, this is pretty fun. Then as I was, like, getting deep into the Swift community, I found out of Adapter. This is for student contest. I was like, this sounds fun. Let me try competing. Then I made an app that was basically a rhythm trainer with computer vision on the iPad. And it won the first year. And afterwards maybe let me trade in this Apple internship stuff. And then that happened to work out, and now I've just been writing Swift code for, like, close to five years.

[Host 2]

Oh, that's awesome.

[Carlos Bandera]

Did you.

[Host 2]

How did you get into tech in the first place? Was Apple development the first time you programmed, or did you do programming and, like, tech stuff before that already?

[Carlos Bandera]

Are you familiar with the TV show Silicon Valley?

[Host 2]

Yeah. Yes. And you're like, those seem like reasonable people to hang out with. I want to do that.

[Carlos Bandera]

Yeah.

[Kai]

I'm not sure if that was the message they tried to convey.

[Carlos Bandera]

I was like 15 or 14. I was in high school. I'm like, cool, so you can make an app and it can get, like, super famous and popular. I can do this too. So I went to Passera, got like a $10 Java programming course. I'm in a very web app. But I was like, this is going to get me rich. I'm going to turn into the next few days of this. I didn't work out. I still ended up going to university, but eventually getting to Swift.

[Host 2]

Yeah, that's awesome.

[Kai]

I wonder how many people watched that and, like, were not in tech before they. Before the show started and then ended up going a tech track after. Because I think for us, we were already developers when it came out and. And that was sort of what drew us to watch it in the first place. It's very interesting show.

[Host 2]

It's a fun show and it holds up surprisingly well. There are so many AI references in it that seemed absurd back then, but now it's actually pretty accurate.

[Carlos Bandera]

Yeah, definitely.

[Host 2]

It's pretty fun. And then after that, you also started working on your own apps. What were. What was the first app you published?

[Carlos Bandera]

Right, so the first app I published to the Mac App Store was an app called Reze. It was like an AI job assistant. Like, you gave your resume, did some ats and analyze your keywords of job descriptions and also use Apple's foundation models. Like we practice my interviews.

[Kai]

It did.

[Carlos Bandera]

Okay. It wasn't a great app, but it was a fun project.

[Host 2]

And it was Mac only, right?

[Carlos Bandera]

Yeah, it was Mac only. It was Mac OS 26 only because I used the foundation models.

[Host 2]

How did you launch it? Do you plan for it to be, like, a big thing? Do you prepare your. Do you have, like a press kit? Do you have those kind of things? Or how did you approach or did you mostly build it as a fun thing for yourself and you're like, if anyone likes it, that's cool too.

[Carlos Bandera]

It was definitely the latter. Like, I built it just because I was like applying for jobs at a time and I was just like, trying to get my resume a bit better. I was like, let me try to automate this process. And then over time, just shipped into an app. But then my second app, I feel like I learned a lot from that experience of the App Store review process that I was like, let me put in a lot of effort of the second app. And that's how setu. And I guess we can talk about that later, but that's how SETU came to be.

[Host 2]

We can talk about it right now. How did that come to be?

[Carlos Bandera]

Right. Have you heard of Treehacks before?

[Host 2]

The Hackathon?

[Carlos Bandera]

Yeah, the hackathon. Stanford's Hackathon.

[Host 2]

Yep.

[Carlos Bandera]

Yeah. So I went to Treehacks, like in February this year, and I met some pretty cool people. And my initial idea going in was like, I want to make an AI dj. But our team decided to pivot to something more relatable. So we did like an AI therapist that's like on device on your iPhone. But afterwards I was still like, I really want to be that AI dj. So I was like, I want to sit down and try making that a reality. The more I started working on it, I started showing my friends, doing some test, like, betas, like, the idea just kept getting better and more refined. I'm like, let me make this an indie app that people actually use. Early in March, I posted on r AppleMusic. Got a lot of great feedback, lots of downloads in the first week, but like a quick version 1.1 update and people liked it even more. And I was like, let me continue working on this until it's like a really good app that I can feel satisfied with. And that's what I'm working on.

[Host 2]

How did launch go? Was the launch strategy there Reddit, or did you have other ways to try to get a word out too?

[Carlos Bandera]

Yeah, so Reddit was my most successful strategy. I think I got like a thousand downloads in the first three days from Reddit and then I also went to other platforms. Like I was on my Discord servers, like some slacks, I mean, to like Iris folks and some communities I'm into. I basically told my friends and dear friends to try it out and give me some reviews, but really it was like the best performing avenue.

[Kai]

Who would you say the users are now? Is it people who want to have like that sort of curated music for themselves.

[Carlos Bandera]

Right. So the main market seems to be people that really enjoy listening to music and are very intention of their listening, but have a hard time discovering new music. So it's basically just like Apple Music. Users who don't really like the Apple music discovery algorithm put some more intentional effort into it.

[Kai]

It's surprising. Yeah, it's surprising how different the Apple music algorithms seem to be for discovery compared to something like Spotify. That's always been a thing that I was surprised by how little new music is being surfaced.

[Carlos Bandera]

Yeah, I feel like Apple music is very much like quality focused. Like it has the best voiceless in my opinion. But discovering new music is kind of hard.

[Host 2]

Yeah, no, that sounds really cool. How did the hackathon go?

[Carlos Bandera]

We didn't win, but I did learn a lot. Like, there are lots of founders at the hackathon. I met some folks from YC Y Combinator giving me really good advice that I really used into the set of launch.

[Host 2]

Oh, so the hackathon was fully in person at Stanford?

[Carlos Bandera]

Oh, yeah, it was in person to put everybody out.

[Kai]

So how does that work? So it's a hackathon at the university, but it's not only for people at Stanford.

[Carlos Bandera]

Yeah, I feel like Stanford University has too much money, so we kind of have to find ways to spend it.

[Host 2]

They have so much money.

[Carlos Bandera]

Yeah, it was crazy. My teammates are from, like, Europe.

[Kai]

Okay, okay. So can anyone just apply to participate in that?

[Carlos Bandera]

I feel like the main audience for, like, participating is, like, mainly, like, students and maybe undergrad, like grad students. But also people can apply to be judges and they do for, like, reimbursements and flights covered. Like, my teammates were, like, from Zurich and London. They had their fights covered. My fight was covered from Kansas. And of course, people from the Bay Area just come and also have, like, a lot of investors. I think Elon Musk is, like, human. Human. I don't know what to call whatever his venture capital firm is. Had a price pool of $200,000 for whatever team want to invest in. YC also invested in some companies there. There's a really interesting hackathon.

[Host 2]

Yeah, that's really cool. Yeah. And I think the Stanford campus is also. It's fun to be there. And it's also just a nice campus and I like the area. I think it's a cool. Palo Alto is a fun little town.

[Carlos Bandera]

Yeah. Super pretty of the palm trees and

[Host 2]

the sunsets and yeah, they do have a lot of my $41 billion in endowment money. That can buy you some flights from Zurich. That's super fun. So now back to your app setu. Do you have new features that you're working on now or what's the state of the app and what's your plan for it?

[Carlos Bandera]

Yeah, so I just finished working in version 1.2, but I haven't published it yet because I want to do some press stuff. So I've already submitted to the App Store nominations. I've sent a few different magazines like MacRumors and 9to5Mac just trying to get it out there. Whether not to be featured beats me, but there's no harm in trying to just been polishing the app, putting some accessibility features and whatnot.

[Host 2]

Did you submit the first one to the Apple nomination? Do you go through the Apple nomination process for the version 1.0 as well, or is this the first time you submit the form?

[Carlos Bandera]

This is the first time version 1.0 was not as awesome as the current version. It's just more of like a proof of concept. And like I said version 1.1 very quickly in the first week because I got lots of awesome feedback from the R Apple music community. And version 1.2 is just like that. Feedback and some polish all put together to make the actual proper app of it.

[Host 2]

That's really cool. We'll put a link in the show notes. Do you have a release date for the 1.2?

[Carlos Bandera]

Yeah, I'd say first week of April. Definitely first week of April.

[Kai]

Nice.

[Carlos Bandera]

Maybe three days from now.

[Host 2]

That's soon.

[Carlos Bandera]

Yeah.

[Kai]

So you also mentioned before, just briefly, that you are very into music in general and you've done some radio hosting, is that right?

[Carlos Bandera]

Yeah, that's right. So I used to DJ here in Kansas at KJHK 97 FM and also in Europe at the LEED student radio, which is in Leeds, England. Yeah, yeah. And also in San Francisco at Bayfift fm. I'd forced the city to listen to my music taste every Thursday at midnight.

[Host 2]

That's awesome.

[Kai]

Oh, that's cool. How does that. How do you get into that? How does that work?

[Carlos Bandera]

Well, basically I was like a freshman in college and I was looking for community, like any new student. And I found out that radio station was like recruiting for people. I was like, oh, this sounds fun. Let me sign up. And I just didn't stop. So whenever I moved to a new place during my undergrad, I used to go to like, oh, is there a student radio station? And I would sign up, but for the one in San Francisco is a really strange story. Because it was during my interviews, I had like five minutes left and I asked, like, my interviewer, like, oh, what's your favorite music genre? And he was like, oh, I really into, like international music. I radio host. I was like, oh, that's crazy. I'm also a radio host. That's crazy. Then, like, he invited me to his radio station and I just kind of like DJed there for a few summers.

[Host 2]

What does that sound like? Do you prepare for your. If you have your set, do you have your time? Do you prepare for it? Do you have your sets set by the time you go to the station for the night? Or, like, how does that.

[Kai]

I have no idea how this works.

[Carlos Bandera]

Yeah, so it's a bit like a podcast, but with music in a way. So basically, if I have a lot of free time and a lot of effort that week, I can plan my entire set beforehand, curate it, have some stories that I want to share in between the songs. And my show is based on, like, international music. So I'd like to do some research about the culture and like, maybe what inspired the music. Like, let's say if I'm doing like Japanese math rock, I'd be like, oh, this is why it's trendy and this is why it may be popular in Osaka. Or like, why Tokyo Mavericks is a bit different. And like, I just go there, pre record my show live. Sometimes I can pre record it if I maybe have a conflict at the time and then I can just let it go live to the Internet or the FM studios.

[Host 2]

That is really fun. How do you find all that new music? Because international, especially international music, seems harder.

[Carlos Bandera]

That's a great question. That's. That's why I put Sato.

[Kai]

I feel like that was a layup there, guy.

[Carlos Bandera]

That is a great setup. But before Sergio, I just used to hop around. Like, I'd go on Apple music and just to listen to music, but I'd go on Spotify Discover Weekly because it was pretty good. But actually after YouTube accounts, like, I have my personal YouTube where I get my main content and another kind of swing for music consumption. YouTube recommends me some pretty naughty, crazy stuff sometimes.

[Host 2]

Yeah, I. I do find YouTube's algorithm for music is surprisingly good of surfacing new stuff.

[Kai]

Yeah, it's weird because it's not great for video. In my experience, it's like, video, of course it's good, but it's a bit fanatic. It's like I watch one video about trains and now all of a sudden all videos are recommendations for train videos.

[Host 2]

That's Real. And how many of them did you watch? Half of them.

[Kai]

Sure, I kept watching, but that's not the point.

[Host 2]

Maybe it's just catering to your fanatic nature of getting overly obsessed about one thing for three weeks.

[Kai]

Yeah. So I think one, I did have actually one follow up question. Because you're a student and you're graduating soon. I hear many people who are saying like, hey, if I was a student now or if someone would ask me what to study now, I wouldn't know in the world of AI, because the world is changing all the time and I feel like I never actually heard a student say that this is a problem. So I'm curious, from your perspective, how do you feel like the AI and sort of, I guess the way work looks in the future is changing consistently. How do you feel about that and how do you approach that as a student?

[Carlos Bandera]

I don't know. I feel like it's a very interesting question because if somebody was a freshman right now, I feel like that advice would be super valid about like, I don't know what the future is, but as somebody who like when I started like university, that's like when ChatGPT just launched like in like 2022. So I've kind of had seen the entire arc of like AI being like, oh, it's this cool thing people use for their homework once once in a while. Yeah, Being like, oh, you got to use cloud code, you got to be on Codex, you got to be on 5.3, got to be AI first software engineer. So I feel like I've had the benefit of seeing the entire AI industry develop over time and also see it get adopted in industry. So as a new grad, I feel like you're able to bring in more AI first perspective to engineering. I was also someone who didn't have AI in freshman and sophomore year. I did also get the DSA and lead code handled. I feel like that middle ground is the best spot to be in because I've said to you, I had you in a lot of research papers just about recognition engines and machine learning is just applied statistics. And I felt cheated, but I did have some of my stats classes to help me with that.

[Kai]

Yeah, that makes sense. And I think, I mean, we feel the same because we have been in the industry and now we're using AI tooling a lot for the things we're building and we feel like it's very beneficial for our work because it's the two of us working together on a lot of things and we can just feel like we can focus on the thing where our time is worth being more spent on. So it's been useful for us. Is that something universities or your university has been encouraging you to do, like using those kind of tools?

[Carlos Bandera]

I feel like it really depends on the professor. Some of my professors are just like use AI, write your essays with AI, do whatever you want with AI. Oh really? A lot of professors go out of their way to like make sure that you're not using AI. You're like, okay, I'm going to craft a very specific prompt and question that AI will fail but it's only answerable if you attend class on Tuesday at 5pm I'm like a really big spectrum but in industry, like when I had my internship the first summer and second summer, it is a very big vibe shift.

[Host 2]

Yeah. Do you see in like when you go through interviews and you're applying for jobs, do you see a difference in the interview process as a result of the change in AI and development tooling?

[Carlos Bandera]

I think it really depends on the company. Like I recently had a reach out for Meta and they had a lot of AI questions in the email itself before even interview. It's like how are you using AI? How are you. How do you measure good output of AI? How to use AI in your day to day work? That's an entire essay of a questionnaire of how using AI in work. Other companies, they say please don't use AI for our technical. We want to evaluate you as a human being. And that's like. Yeah. And some companies like AI.

[Kai]

What's that? That's.

[Host 2]

Yeah, that's a bad thing.

[Kai]

How did. Because you built an app for your. For improving your resume. How did the job hunt go?

[Carlos Bandera]

I don't know. That's a great question. Well, I do have some offers lined up but I'm not really sure if it's like a single offer I want to take right now. After being set to as an indie developer, I kind of have more fun with it. I kind of feel worried to join a big tech company because I feel like it's a bit restrictive in what you can do as an individual and how you can use your tools. At the end of the day the objectives raise your holder value and not necessarily be the best product.

[Host 2]

Yep. I mean I can vouch for building stuff yourself. It's good, it's a fun time.

[Carlos Bandera]

It's a lot of work.

[Host 2]

It's probably longer hours than any job you would take but I think it's worth it to give it a shot.

[Kai]

What is the Monetization strategy for Seto. Is it a subscription at the moment?

[Carlos Bandera]

No. I thought SETO is more of an experiment for me to be honest. I just wanted to see because right now I'm serious student. I feel like I have the benefit of just being able to experiment and fail with no ramifications. Just kind of wanted like see, can I build an app? Can I have it like get users, can it be downloaded and can it scale? And that's why I try to make sure it's on device and like someone self sufficient just so I don't really have any expenses to do. But afterwards I want to build like another app. I think version 1.2 is going to be my last version of set 2 and I want to work on something that's more monetizable and like more scalable.

[Host 2]

Do you have an idea already for what are you going to build next?

[Carlos Bandera]

I have a few ideas, some a bit more radio focused, but it's too early to see.

[Host 2]

I also had a follow up for the student situation. What is the general vibe with other students? Do most people fully embrace AI tooling or there's some people that are really against it. What's the general vibe in class I

[Carlos Bandera]

feel like everybody in class has used AI, but I feel like the extent to which they use it really depends. Some people are about the same level of AI proficiency as my mom. Whereby. What's this feature? Gemini 3.5. That sounds cool. I still use ChatGPT just like the text box. I have some friends who are building custom MCP servers for the cloud code integration for cloud code, Codex and other platforms. It's really a spectrum of some people are really into it, some people are kind of into it, but I feel like that's normal for most schools. You have some pretty cool people, some pretty casual people.

[Host 2]

Where do you fall on that spectrum?

[Carlos Bandera]

I don't know. I feel like somewhere closer to hardcore. But not as hardcore as my other friends. Maybe I let them experiment with the breaking in for a month and see what sticks. And then I copy your homework. I'm like, okay, yeah sure, I'll try it too.

[Host 2]

I think that's a smart spot to be. You get to see what other people are doing. So you're staying up to date, but you don't waste your time on the things that don't work. It's pretty smart. I like it.

[Carlos Bandera]

It's been working so far.

[Host 2]

So now about your talk at Deep Dish. What's your talk about? What's your talk title?

[Carlos Bandera]

Yeah, so I'm giving a talk on extra Saucy Swift audio APIs. So last year Apple released a new framework called Speech Analyzer which replaces the Speech recognizer from like 2016 and iOS 10. And it's like a pretty cool framework. And I'm going to be talking about that as well as some other open source alternatives that people have been working on because I feel like that Speech ML Field has really been developing especially Napa ecosystem systems.

[Host 2]

That's really cool. Is that a iOS 26 only API? Is that the new one?

[Carlos Bandera]

I'm not sure if it's iOS 26 exclusive, but it did come out like last year.

[Kai]

Did you decide on the Talk before releasing the app or has this sort of been a journey where you explored APIs for Seto the same time right as preparing the talk?

[Carlos Bandera]

That's a great question. As you started working on like the Talk and technology before I like Speech Analyzer even came out. And I was using more open source methods like Stripper, Onyx and Fluid Audio. Those are like some open source APIs that developers can clutch from GitHub on using their apps. Right. And it all started cause I was really into podcasts around 2024 and I really enjoyed Spatial Audio. But I was like, it's kind of weird that you watch like a podcast and you see two people sitting side by side, but it starts doing mono. It's like in stereo when you have a group of people.

[Kai]

Oh.

[Carlos Bandera]

I was like, how can we use computers and algorithms to specialize the content automatically? Started doing some research. Eventually I got really deep into Speech ML weeds, started doing some stuff like embeddings, speech generation and et cetera. Made some open source contributions that when Apple Speech Analyzer came, I'm like, wow, this was great. I wish this existed two years ago. And so I just want to give the developers a quick leapfrog into the cutting edge of Apple's AI and Speech ML stuff.

[Host 2]

That's very cool.

[Kai]

Yeah. And how is it to work with? Do you feel like, is it well, is it APIs that are well documented, does it work?

[Host 2]

Do they work?

[Kai]

Do they work?

[Carlos Bandera]

I feel like the best part is that it's very swifty now because I feel like the 2016 API was very much a product of its time. Like you had to do so much work just for you to get started. But now if the new Speech Analyzer, you write like five lines of Swift code and it's done and you have like output, you have transcriptions ready to go. So it really likes to iterate faster and now if you have like an AI agent writing your code for you, it's easier to review code so in five lines versus having like a view controller that's also doing some pretty strange stuff.

[Kai]

And what's the state of your talk? Are you almost done or you're thinking about it?

[Carlos Bandera]

I would say it's almost done. I've been kind of thrown about it. I wasn't sure if it should be more technical or like more high level of it. I think I want to pivot to make it more high level and just make like a starter project, maybe like a companion article people can read into to like get into the docs and like details of the project because there's a lot of content that you want to get through, like the Apple stuff, the open source alternatives and how you can connect it with AI and on device models and stuff like that. So I feel like right now I'm just doing some polish and like some iterations on it.

[Host 2]

How does the speech analyzer compare to the, I assume bigger models that you would run in a cloud? Does it have features that you would expect from a speech analyzer otherwise, like pitch and could you, could you detect like nervousness in a speaker's voice? Does speech analyzer do those kind of things or is it more focused on semantic and transcription and those kind of things? Semantic understanding and transcription I guess.

[Carlos Bandera]

Right. I feel like it's important to like decouple speech analyzer from the entire pipeline you do from the text process inside or like maybe the voice analysis like speech analyzer, you can ask it like do some transcription for me and I'll give you like the text output and it's all on device and it's private so you can do it for as much as you want. But also if you want you can ask it. Can you also give me the pitch and the variance of the person's talking and whatnot. And now it's up to you as a developer to interpret those raw numbers into something useful. Now if open source models, you can now maybe get an emotion classifier, not for audio specifically, and put it on top of it to process that audio file. But Split analytics is more focused on the getting started part. And the way Apple has designed it this time is more modular. So you have a speech analyzer at the high level and now it has analysis session, then you can have a transcription session under that or a dictation session under activity session under it. And now you configure it however you want your results to come.

[Host 2]

That's Cool. Does that mean you could also. Would that also allow you to essentially put different pipelines together with some open source components and speech analyzer and have them work together?

[Carlos Bandera]

Well, yeah, I feel like Neri's plane may talk. Yeah. But yeah, that's essentially like the end goal. Like for example, you're a speech analyzer to maybe get a transcription. And now if you use an open source like speech organization model that lets you label like the conversation based on who's talking, it's like, let's say we have like a conversation of the few of us, right. Be able to be like, oh, at this section person A is talking and at this person B and this person is person C. So it gives you that label. And maybe if you have like an AI pizza delivery app, you can be like, oh, person A wants a pizza of barbecue chicken, person B likes pizza. And then you can use that with, I don't know, some ordering API or something like that.

[Host 2]

Yeah. What's the speed of the speech analyzer? Does it work real time or is it fast in real time?

[Carlos Bandera]

If you've used Apple's feature for the voice recording, like on the notes app, it's basically the same API app was using for that feature. And also like in the podcast and Funko features and AI summarization stuff, they make a big point in a dub dub video that this is the same API we're using our stuff and you developers can use it too. And when we update it, we also get updates.

[Host 2]

Yeah, it's fun. I feel like Apple made quite a big shift in their AI tooling. It felt like there was a bit of a rough period, but I do think the current approach is a lot more modular in general, not just for speech analyzer, but I think in general for the AI tooling. I think that feels like they're on a better track now, which is. Yeah, goodness.

[Carlos Bandera]

Yeah.

[Host 2]

Awesome. I think that brings us to our final and hardest question.

[Carlos Bandera]

Oh, wow.

[Kai]

You always get people. Oh, I'm the one saying it. I'm the bad one.

[Host 2]

Yeah, you're a bad one.

[Kai]

Yeah. Who is a person or a company or an entity or anything in the world really that inspires you.

[Carlos Bandera]

Oh, that's really hard. I don't know, I guess I'd have to say Najabis. Najabis was a hip hop producer from Japan in the early 2000s. This anime called Samurai Champloo. That's kind of like their philosophy they had with their music. It was like very cross cultural directed. People from different countries made some cool music Put in some pretty cool projects. And I just want to have that attitude as I go through industry and whatnot. Some pretty cool stuff with some pretty cool people.

[Host 2]

And it's like, yeah, that's also a very unique answer we've not had. You don't have to be worried about clashing with someone else. I think that's. That's pretty niche. How did you come across that work? Was that part of your music journey or your manga anime journey? How did you.

[Carlos Bandera]

Yeah, I would say it's actually how I got the journey started. So I had a friend back in high school who was like, oh, you should watch Summer Champliu. So just really good summer anime. And I was like, okay, I'll check it out. This friend of mine was like a music producer in his free time. So it was like, yeah, it was the soundtrack. Was it by New Jabes? Who's this? And then I learned about the lore about it. I was like, this is pretty cool. And then I got more into anime and manga. I found out the director who did Samurai Champloo also did Cowboy Bebop, which is like a cult classic. And I was like, wow, this is kind of fun. And I just started going deeper from there. I made jazz beats for it for some time. I don't use music anymore, but that's. That's like history.

[Kai]

Do you. Do you play any instruments as well?

[Carlos Bandera]

Yeah, I played a guitar. At least I try to.

[Host 2]

Did you consider music as an alternative to tech, or did you always feel like combining the two is the sweet spot for you?

[Carlos Bandera]

Well, I. Firstly, I feel like tech's always been fun for me ever since, like, growing up and whatnot. And I asked my mom maybe I should do a music degree instead of a tech degree, and she said no. And so I got a degree in computer science because my mom was like, you can do the music stuff for fun.

[Host 2]

I think that smart advice not going to discourage anyone from music degrees, but I think tech as a base is a pretty good base level degree from which you can then explore all your different things and combine them. Right. I feel like that's a very unique thing about technology. If you're into film and tech, you can use your tech skills to do things in film or music or photography or art or whatever you want. I think it's a really good kind of binding box to whatever other interests you have, which then makes you be able to uniquely build something in that space. I think that's a very, very fun part of tech that it's so varied and flexible to be molded into something that you're passionate about.

[Carlos Bandera]

Yeah, definitely. I feel like that's where the magic happens when someone should be passionate about one of our artsy things and some tech and I think some pre projects in between.

[Host 2]

Yeah. Awesome. That was really fun. Thanks for joining us on the show.

[Carlos Bandera]

Yeah.

[Kai]

Thank you so much.

[Carlos Bandera]

Thanks. It's been a pleasure chatting.

[Host 2]

Yeah. Good luck with your final polishes on your. On your talk and looking forward to seeing it in Chicago.

[Carlos Bandera]

Looking forward to seeing you, too, in Chicago.

[Host 2]

Awesome. Perfect.

[Kai]

All right.

[Carlos Bandera]

Bye. Bye.