AccessWorld Podcast, Episode 40: Tackling that Inbox with Lumin

Episode Notes

An interview with Sriram Ganesh, CEO and developer of the AI app Lumin

In this episode of AccessWorld, a podcast on digital inclusion and accessibility, Aaron and Tony sit down with Sriram Ganesh, CEO and developer of the AI app Lumin. Lumin offers a voice controlled solution for managing your inbox and calendar. It offers a hands-free way to tackle that cluttered inbox, and was born out of Sriram's own lived needs for greater accessibility after losing his sight to cone-rod dystrophy.

Sriram shares how his love for video games and past success as an AI developer led him to creating such a user friendly approach toward bringing together multiple AI models into an app that focuses on making the user experience as simple as possible. (Note: there are a couple moments of brief silence as they demo the app in-between AI responses, so don't worry, it's not your headphones.)

You can learn more about the app at www.trylumin.ai.

Aaron also shares what's in the latest issue of AccessWorld Magazine, which dropped last week. You can check it out at www.afb.org/AW.

AccessWorld is a production of AFB Studios at the American Foundation for the Blind (AFB). Feel free to like and subscribe to the podcast, and even consider visiting our site to make a donation to support our work creating a world of endless possibilities for people who are blind or have low vision.

This podcast is powered by Pinecast.

AccessWorld Podcast, Episode 40 Transcript:

Intro: AFB. You're listening to AccessWorld, a podcast on digital inclusion and accessibility. AccessWorld is a production of AFB Studios at the American Foundation for the Blind. To learn more about AccessWorld, visit us online at afb.org/aw.

Tony Stephens: And welcome back, everybody, to AccessWorld, a podcast on digital inclusion and accessibility. I'm Tony Stevens, leading Communications at the American Foundation for the Blind, your co-host. And joined with me today, as always, is our fearless editor-in-chief of AccessWorld Magazine, a quarterly publication that shares the same name and actually just dropped the other day for the summer into fall quarter, Mr. Aaron Preece. Hey, Aaron, how are you again, sir?

Aaron Preece: I'm doing fantastic, Tony. How are you doing today?

Tony Stephens: I'm doing well, but I was coaxing the nerves of an anxious guide dog, German Shepherd, who they come across as so intimidating, but they're such babies when you know. So, you know, a little, a little therapy and lots of pets after we finish recording this. But we are excited today to have the CEO of LuminAID, a creator of an app that is pretty interesting. If you, those that had a chance to listen to our episode last week, or 2 weeks ago it was, when we had the AI research team on here from AFB talking about the assistive technology research report that they pushed out. We're excited actually in that conversation of the breakthroughs with AI and things that are being developed in assistive tech to have a developer himself with lived experience who will be sharing some information about the app and give us a sense of where this technology is being best employed and deployed. And so we'll be talking with our guest in a second. But in the meantime, though, Aaron, we just dropped AccessWorld a few days ago. Is that right?

Aaron Preece: Yes, it is live on the website.

Tony Stephens: Excellent. Yeah.

Aaron Preece: So this issue, we have 5 articles. Uh, we have a couple of product reviews. We have the EasyReader suite of apps from Steve Kelly, and that's a sort of a program suite that tries to make computers easier and more straightforward to use for people that have low vision or seniors or people new to computers. Be My Eyes is now on Mac, and Janet Ingham reviewed that. If people remember back whenever we talked about Viewpoint back last summer already, which is crazy time-wise, it lets you like take snapshots of your screen and get them described by Be My Eyes and Be My AI. In this case, it's very Be My AI focused, even though it's Be My Eyes. So it's their desktop application for Mac. It sounds very efficient to use. It sounded pretty cool. So there's a review of that. We have an article I put together on PDF accessibility and like PDF pitfalls. There's a lot of pitfalls you can fall into if you're trying— it's like the old video game, whenever you're trying to make a PDF accessible. They can be great if they're done properly, but there's a lot of things to consider. So we went through that. There's resources on more details and more technical details of how you can do that in the article.

And then I introduced a new category this issue. I call it Editor's Experiences. There's a lot of time, there's things that we talk about on here or just stuff I do in my day-to-day life accessibility-wise or technology-wise or things I'm kind of diving into that I think AccessWorld readers would find interesting. And so this time I revisited Dungeons Dragons back 7 years ago, basically almost 7 years ago to the day or to the month. I talked about Dungeons Dragons and accessibility because the edition at that time had just become accessible from a book access standpoint. And I had been playing for a couple months. I've since gone through a full 0 to max level campaign that stretched over 6 years. I currently play in 3 separate Dungeons Dragons campaigns that are all going on over a year now after that long campaign. And so I wanted to revisit Dungeons and Dragons and tabletop role-playing games, they're one of my favorite hobbies and I wanted to revisit them because they're very accessible and there's some things, basically just kind of see what's changed.

And luckily things have kind of, they were already pretty good back in 2019, but they've really improved since. And there's more accessible dice, there's more books available, the field has expanded and that hobby is just continuing to grow. And then finally, I, in the last couple months, I have really dove into vibe coding for accessibility. And I think we talked about it on one of our podcasts when we talked about ACB, but one of my favorite sessions there was about vibe coding. If something's not accessible or you need something that's a very niche kind of program that, that's just not available, you can use vibe coding now to make an accessible version of something that might not exist otherwise, or you can Vibe code, like a patch to an existing piece of software potentially. So I talked about the utilities I made and I talked about the game accessibility mod that I'm currently working on that's gone through multiple iterations. So that is the fall 2026 issue.

Tony Stephens: Excellent. Well, thanks so much for sharing that. And for folks that want to check out the latest issue of AccessWorld, you can go to afb.org/aw and check out the latest issue along with 26 years of back issues, all for free. And you know, you're talking about the video games, which is exciting. I want to get into some video game conversations with our guest as well too, because I hear it was maybe a point of passion as well in the development space. But we are excited to have Sriram Ganesh joining us, who's running a new app called Lumin. And I think we're going to get a chance to check that out. But first, let's meet the developer himself leading the effort to make our world more accessible at the hands of AI. As we work to expand new ways that chatbots can make the world easy for us, particularly those who are blind or low vision that oftentimes get caught in quagmires themselves of trying to navigate things using screen readers and other access tech. But like I said, fitting with our access tech issue that came out a couple of weeks ago. But, but Shyamam, greetings, salutations, sir.

Sriram Ganesh: Thank you so much, Tony. It's great to be here.

Tony Stephens: Share with us a little bit just about the app at the top. And I mean, what better way maybe to jump right in, but let's literally get a chance to hear it, maybe right out of the gate. But what is Lumin?

Sriram Ganesh: Yeah, I founded Lumin because I'm an engineer. I like building things. And that's what I've been doing ever since I started working. And AI has been a space that I've been in since the early days when AI started working, which is almost since 2010. And so it's great to see a lot of these technologies mature today. But Lumin really started because of my own experience with vision loss. I have a progressive central vision loss condition called cone-rod dystrophy. Simple terms, it affects my central vision. Reading documents, texts clearly essentially have become harder for me. And as I started looking for solutions that I would really want to use to really help bring back my agency that I felt like I was losing, some of the independence that I feel like I was losing, and even like, you know, that affects the confidence as well, right? Like, it affects so many parts of your life.

After a while, I was just motivated to see how can I contribute to the space. And so about 6 months ago, I put together a team, and we are a team of 4 now, and we started building Lumin. Lumin started with an inspiration that for people who are losing their vision or people who either can't or don't want to stare at a screen and get stuff done, we need a better way. And we started with an idea of building a voice assistant. Most people think of voice assistants assistants like Siri or Alexa that do a few things, but we wanted to take that to a whole new level. A voice assistant that sits on your phone that you can completely depend on and do pretty complex tasks like email and calendar to begin with. I'd love to talk to you more about this, but I think in order to kind of anchor the audience towards what we're really talking about, I'd love to show a quick demo. What the audience will hear is coming straight out of Lumin, which is running on my iPhone. The screen just has one big button, which is Start Lumin. And when I tap on that, I'm basically in a conversation with Lumin. And this is a real demo working out of my actual inbox. So, you know, let's brace ourselves.

Lumin: Good afternoon.

Sriram Ganesh: Hey, Lumin, can you tell me what's new in my inbox, please?

Lumin: 15 unread emails today across your connected inboxes, prioritized by what needs attention. 1, Mikhail, a meeting invite for a compose deep dive this afternoon. 2, Steven Poirier, product feedback saying Lumin changed draft wording and lost their draft. 3, Pat Quinn, investor presentation schedule and room details for tomorrow. 4, Anthony Stevens, a note ahead of today's podcast. 5, Andre, thoughts on adding external link support? Go one by one or open the next batch?

Sriram Ganesh: Can you, can you open what Pat is saying on the first one?

Lumin: Do you mean Pat on 3 or the first email?

Sriram Ganesh: Actually, let's go with Pat on 3 first. That sounds more interesting.

Lumin: RE: They say they sent calendar invites for tomorrow's investor presentation and a rehearsal at South Park Commons. They also attached a photo of the stage with the moderator and asked you to let them know if it all looks good. Want the image summarized?

Sriram Ganesh: Yeah, because, you know, I can't look at the image right now. Can you summarize the image and give me a sense of what that is so that I can respond to Pat?

Lumin: Yeah, it looks like a polished interview-style stage setup. There are 2 dark blue armchairs on a low platform facing the audience with a small round table between them and a tablet on top. Behind them is a large mural-style backdrop with line drawings of people and a big wood logo centered between the chairs. The lighting looks warm and even with plants around the edges, so overall it sounds professional and well put together. Want me to draft a quick reply to Pat saying it looks good?

Sriram Ganesh: Yeah, please. Can you do that for me? Actually looks pretty good.

Lumin: This is a reply to Pat Quinn. Wrote, looks pretty good to me. Thanks for sending this over and for making sure there'll be water on stage. Ready to send?

Sriram Ganesh: Yep.

Lumin: That's sent. Coming back to your emails. One—

Sriram Ganesh: Can we—can you go to the next batch? I want to see what else is there.

Lumin: Here's the next batch. 6, Mikhail. Good news on improving user experience ratings since Tuesday. 7, Pat Quinn. An AFB report with a PDF attached for today's podcast. 8, Pat Quinn. A calendar invite for tomorrow's investor presentation rehearsal. 9, Pat Quinn. A calendar invite for tomorrow's SPC investor presentation. 10, Mercury. A product update about insights. Go one by one or open the next batch?

Sriram Ganesh: Um, let's, let's go one by one.

Lumin: 6, Mikhail. They say they're seeing positive trends in user experience ratings since Tuesday. Want more details?

Sriram Ganesh: Uh, you know what, I'll take a look at this later. I just want to move it out of my inbox. Can you archive this for me?

Lumin: 7. Pat Quinn. They resurfaced the latest AFB report for today's podcast and attached a PDF. Want the details?

Sriram Ganesh: You know what, Lumin? I hate reading PDFs on my computer. Please help summarize this for me, please.

Lumin: AFB report on how AI is being used as assistive technology for disabled people based on a survey of 1,735 people in 2025. The big themes are visual description, automated captions, and AAC. Adoption is high, especially among blind and low-vision users.

Sriram Ganesh: You know what, Lumin? I was, I was talking to someone, and they mentioned that this report had some insights about healthcare. Can you find that for me?

Tony Stephens: I have an enormous grin across my face.

Lumin: Yes, the healthcare-specific points are about safety, clinical communication, and privacy. The strongest example is a blind or low-vision user who used AI to read medication instructions and got a dangerous wrong result telling them to chew 4 tablets twice a day for a topical medicine. The report uses that to show why healthcare is a high-risk setting. Users may depend on AI for medical instructions they can't easily verify on their own. It also points to deaf or hard-of-hearing users—

Sriram Ganesh: You know what, Lumin, I would love to keep chatting with you, but I'm in the middle of a podcast, so I have to stop right now. Would you happen to find an email from, uh, uh, is it, is it an email from Anthony today? I think he might have sent me something. Can you check?

Lumin: Yes, two from Anthony today. One is from Anthony Stevens about today's podcast.

Sriram Ganesh: Yeah, can you please open this one and tell me what they're saying?

Lumin: Which one? The podcast note from Anthony Stevens or the app test email from the—Anthony Stevens says they're looking forward to having you on the AccessWorld podcast today. They also said they enjoy testing the app and can't wait to see how it works in real time. Want me to draft a quick reply?

Sriram Ganesh: Yep. Say hi to him and firstly say thank you so much for inviting me to the podcast. I'm really looking forward to the conversation. And say, I'm like, I'd love to chat about everything accessibility and, you know, and about Lumin. Looking forward to it. Add a smiley face at the end of it.

Lumin: This is a reply to Anthony Stevens. I wrote, hi, Tony. Thank you so much for inviting me to the podcast. I'm really looking forward to the conversation. I'd love to chat about accessibility and Lumin. Looking forward to it. Smiley face. Best, Sriram. Ready to send?

Sriram Ganesh: Yes.

Tony Stephens: All right, you can just drop the mic now and walk away.

Sriram Ganesh: Lumin, you know what? I gotta go. We can't keep chatting. So I'm actually on a podcast right now, Lumin. So I gotta go. Can you end the call for me, please?

Lumin: You got it. Good luck on the podcast. Talk soon.

Tony Stephens: That was cool.

Aaron Preece: Very cool.

Tony Stephens: Yeah.

Aaron Preece: I'm impressed with that flexibility. Is that, I assume, connected to like a backend, like a forerunner? What would you call it? A frontier model, forerunner model?

Sriram Ganesh: Yes. So, so like in the backend, the way to kind of think about it is we have, we have like this orchestration So Lumin's backend is an orchestration of different models, which makes sure that it's fast, it's capable, in which case, like, it actually gets the emails right. An example of that was Tony's email, but it also has to be reliable, right?

Aaron Preece: Mm-hmm.

Sriram Ganesh: And so our orchestration in the backend works with a bunch of different models to make sure that Lumin is all the 3, which is really the innovation that we bring to the table. If you look at something like a Siri or an Alexa, they are really fast. they're quite reliable, but they're not very capable. They don't do many things. And so, the biggest trade-off in what we are building is you can either get things to be— you can get things to be 2 out of the 3 here, but it's hard to be both capable, fast, and reliable. And that's the magic of what we do on the backend at Lumin.

Tony Stephens: And that's magic. Like, I know it's not magic per se, but I mean, you know, I was talking with someone earlier about the app. and like pitching it to you, say, yeah, I'm checking this app out, you should check it out too, on a call a little earlier. And, you know, the thing that's impressed me out of the gate was the UI for it. It was very simple to jump right into it and start using it. But the robustness of the AI chatbot, it's so much more robust than I have my other chatbots. I use Claude and OpenAI, you know, ChatGPT. And they, you know, they've gotten to know me and they can answer questions. Aaron and I have talked a lot about this on the podcast, the way our bots are sort of, you know, learning along with us. But, but just the functionality that you're able to do, like jumping into our report, which is like a 28-page report— it wasn't a, a one-page summary.

Sriram Ganesh: Yeah.

Tony Stephens: That, that, you know, your, your, your Lumin was, was going through, and just to pull those stats and present it in such an informative way— I mean, it— yeah, I'm, I'm still sort of speechless in a sense that, that, you know, I've been playing with it since I downloaded it, but not to that. I didn't realize it had this kind of horsepower.

Sriram Ganesh: It's actually the biggest challenge for us. It's also one of the reasons why we love doing podcasts like this is we need to show people the abilities, right? It's like audio is like a dark room. It's not like when you look at a visual interface, people have lots of buttons and lots of options. The negative of that is that it's just a cognitive overload. Most people don't want to be tapping a bunch of buttons and like going through rooms and other things to just get something simple. But the beauty of that is, you know what's hap— you know what it— what the product does, because if the UI says that you can do something, you can just do it. With Lumin, the positive is that is the minimalism, which is it takes 10 seconds for someone to learn how to talk to Lumin. And if you don't know something, you just ask Lumin. Like, that's it.

And Lumin doesn't pretend to do 1,000 things. It's not ChatGPT or Claude. You can't ask it about anything in the world. It's very focused on the task at hand, which is I want to help the user get through their emails as best as possible. And as part of that is to go to the last mile, which is if there's an attachment and you can't read it, then the job's not done. Make sure we can read the attachment, right? Like, so today we support images. You saw the one that I replied to Pat about where he took a photo, and this is my new way of exploring photos today, which is he just takes photos and sends it through Lumin, and I just ask it to describe it in the level of detail, and it looks pretty good. It's pretty good. Lumin can also do PDFs, but we can go further. We can do Excel sheets, which is my absolute nightmare. If an Excel sheet opens up in front of me, it is definitely anxiety-inducing. And with Lumin, I just go, hey, Lumin, I have this budget sheet in front of me. How much money have I spent in Uber this month? And it just adds it up and tells it to you. And I'm like, that's accessibility for me. I should just be able to ask questions and get the answers.

Tony Stephens: Again, I mean, we talk so much about UI in the accessibility sense of meaning, how's my screen reader going over the page, and is it finding the things? But this is a user experience, a UX experience that's really just— it is like having someone right by your side, like an executive assistant that's following you around and just is there for the ready, you know?

Sriram Ganesh: Yeah.

Tony Stephens: And yeah, well, let's, let's, let's pivot a little and jump into your story, right? Because what, what got you to think of this? What was some of the things that, that seeded you to walk down a path into the AI space? But just to do just what you said, in a sense, it's not cluttered with buttons. It's not, you know, when digital broadcasting came out and there was a huge thing, this was like 30 years ago, 26 or 27 years ago, when, you know, AOL was the first to sort of purchase up CNN. And then they threw everything on the screen that seemed like America Online screen. Like it was like, it was an overdose. Like, and the nature of the digital world was like, how much can we put into this, this kit, you know, the kitchen sink and everything else to try to give people versus a streamlined— what this does is very simple user experience. But, but talk a little bit about, if you don't mind, like how you moved into this space and got you to this mindset of such a clear and distinct path to create something that, that I'm already thinking of a dozen ways that it's going to be usable for me in my day-to-day workflow of getting hundreds of emails day. But where did this idea come from? And more about, you know, what got you to that point?

Sriram Ganesh: Yeah, you know, without boring you and the audience on too much of my life story, but just to give you kind of the pivotal moments of kind of what happened, like, I, you know, the first time I saw a computer is when I— well, the first thing I saw on a computer was a video game. And it was just this— and I still remember, I was like 12 years old living in India, and it blew my mind. It was the most magical thing that I'd seen. And I've also been a really big fan of Arthur Clarke, who has this incredible line that when a sufficiently advanced technology almost seems indistinguishable from magic. And that kind of concept has stuck in my head, which is I'm interested in, like, I just get attracted to technologies and products that feel like magic to the user. It's like you said, it's obviously not magic. It's like science and technology, but the user doesn't care about it. And the user shouldn't care about it. The user should feel it is magic. And that's the whole kind of delight in this whole thing.

And my sort of love for video games really transformed into me wanting to learn how to build video games. Because I wanted to be one of those guys who could build these things. And that was the inspiration. And that led me to learn the math and the programming required to actually build video games that are super fun. That experience from video games took me into my previous startup experience. I started my first company, which worked on technologies like augmented and virtual reality, which is let's put people in immersive environments to do different things, right? Whether it's entertainment, whether it is to visualize furniture in your living room. Uh, uh, what if, what if you could see furniture before you actually bought it in your living room, right? What if I could like show you whether the couch that you want to buy online actually fits in your room? What if I can show you how the colors look like before you spend $800, $1,000 on buying something? Those are the kind of use cases that we started my previous company with that, uh, grew, and eventually our largest use case became education. And the basic concept was that kids are not motivated by boring textbooks anymore. The problem with education was purely motivation, which is that we don't have to force kids to play video games, then why do we have to force them to learn? It's because the medium of learning is honestly boring. That's just it.

And so our job was, how can we make learning so much fun and feel like a video game that you, uh, you almost jump into it, uh, rather than push it away. That experience— and because of that experience taught me a lot. I worked with, uh, pre-kindergarten kids and built products with Disney to kind of create these immersive games that taught them math. Uh, and what that taught me was if you make something fun and exciting for somebody, they will do it. Like, that's the motivation they need. They, they need things to be easy and they need to be accessible, which is the same problem I was solving for kids. You need like large buttons when you're working with 4-year-old kids because you want it to be easy to tap on. Even though I never had accessibility challenges back then, I was already learning how to build accessible software for kids.

In parallel is when I was diagnosed with my low vision condition, which is cone-rod dystrophy. This was obviously a huge shock to me because I suddenly woke up one day and I was like, why is the text blurry? I don't understand. I wear glasses, it should not be blurry. When I went and met a few retinal specialists, they told me that I have this condition and that it's progressive. And there is no fix. Because my first question was, hey, listen, I don't know anything about this. Is there a pill that I can take? Like, what do I do? Like, what do I do to kind of fix this? And it took me a little bit of time to realize that, okay, I unfortunately have to live with this condition. Because medicine has not progressed so far that I can just go to a doctor and get some surgery or get it fixed. That led me into understanding what are the technologies and products available for people like me. When I saw those products and I spoke to lots of users, what I realized was those products are very helpful, but I personally found them to be challenging to adopt. On the other hand, I also felt like with my background and my experience, I should take a fresh shot at it. Let's create a new experiment. Let's try and do something that has not been done in this space. If we fail, we fail, but if we don't try, we never know. And so around me, I constantly saw users using VoiceOver and speech-to-text tools and a whole lot of these other tools. And I was like, what's going on there? Why are they using these tools? And we realized that when one of your senses starts deteriorating, which is your vision, your other senses become your better friends. And audio had clearly become a friend of users within this community, but I just felt like the audio experience had not been taken to the level at which it should be.

And that was the founding story of Lumin, which is we took, you know, I still remember I was talking to someone from the Lighthouse for the Blind as I was trying to learn about the different conditions our users are facing. And I was emailing them and they were emailing me, and both of us have low vision, and the struggle was just, It was just like too much. I was tired of just zooming in. The moment you get a long email, for me it was game over. Like I had to like zoom in to like and go word by word and it would take me like 25 minutes. And I would, when I wanted to respond to it, I would be so nervous. I'm like, I don't know what mistakes I'm making. Does it, is the formatting bad? Am I making spelling mistakes? How does the other person perceive me? Right? You care about all these things. You want your emails to properly represent you. You don't want it to look like, you don't want it to look like like, you know, I've been inattentive. I've not been inattentive. I just have an accessibility problem. And so the question was, what if we can devise an agent that starts by doing your email and calendar that you can talk to like it was a real person? So we want to mimic a real assistant. If you were someone who's very wealthy and you have an assistant that you can call up on your phone and get stuff done, How can we take that experience without office hours? You can talk to Lumin whenever you want to. Lumin doesn't lose patience. Lumin doesn't need to eat or sleep, but you can call it anytime. It is accessible and it should help you get these things done. That was the core inspiration 6 months ago.

Tony Stephens: Wow. 6 months ago?

Sriram Ganesh: Yeah.

Tony Stephens: Like, not that long. I mean, it's— I mean, it's a— it's a— the space is changing so much. I love hearing about the user I mean, as a, as a video game user, you know what experience is, right? You know the experience that got you into development and software development. But then just the, the, when you were talking about like the kids and now yourself with lived experience, and that's one of the things I think that's great about the app is the user experience of it is it is very, it's even to the point of your tutorial at the beginning, right? For those that download the app and go through the tutorial, it's It reminds me somewhat of like starting a video game. It's funny you mentioned that, but it's like, yeah, we're gonna talk you through how this thing works and go to town with it.

Sriram Ganesh: That is exactly how we did it. And the reason why, if you look at, there are so many attributes of this that I have taken from across my career. If you look at how we do the onboarding, where it's almost like, I'm like, it's like, because when users come in, you know, our community has been sold many dreams and, And when users come in, there is a level of nervousness. And so we try to almost like hold your hand and walk you to the place where you can be by yourself, right? Like, and we walk you through it step by step. It's like, hey, let me just guide you. Like, don't worry about it. You, you, why don't you sit down and relax and I will walk you through it. And once you've sent a couple of your emails, that's when the users go, oh, wow. Okay. So that's how I can, I can get rid of my training wheels almost. Like, thank you for the training wheels. I don't need them anymore. let me start driving this bicycle myself. And that is our goal. And we don't wanna just throw you into the deep end. We wanna kind of walk you through it. If you look at other aspects of the product where the pro— there's a reason why the, you know, when Lumin's trying to fetch your emails, there's a reason why we have boops. And some users who might be sighted might actually think of that and go, that's annoying. Like, why do you have constant boops? The reason why we have constant boops is that the user does not know why they're left in silence. Our user can't see, and so they need an indication. They need an audio indication of what's going on.

Similarly, when you send an email, why do we do the ka-ching? Because in a noisy environment, you want to be very certain of important actions. You want to make sure it's actually gone, because if you wanted to send an email to someone important and something happened and you didn't know that it You're just going to walk around, you're going to be walking around being nervous. And so all of these, and the other aspect of it is why are there so less buttons and why are these buttons so big? Part of it, of course, is we want it to be an eye-free and hands-free experience as much as possible. But a lot of that also learning comes from how we design products for children who are in pre-kindergarten. which is we make it incredibly easy to focus on a few things and we just don't clutter the screen. So, you're absolutely right. Like, Lumin's design is very much inspired by building an audio-only guided video game. It just so happens that it does work for you.

Tony Stephens: Yeah. Well, instead of you know, uh, making a pizza in my kids' Roblox game or fighting in another game or Fortnite. This is, this is checking my email. Like, this is a grown-up, user-friendly experience for really what for many of us in, in, in the world, uh, get overwhelmed with when I open my inbox sometimes.

Aaron Preece: Yeah, I'm, and I was thinking too, it's, it feels like the kind of thing that say someone's completely new to using their iPhone or, or and they wanna be able to check their email, even if they're not as tech savvy, they can still use this. But if you are tech savvy, it still has a lot of applications. I know I get tons of like spam email and stuff. And so the inbox fills up quick and it's the idea that you can just talk to someone and like find the most important things and that this intelligent agent in the background is able to prioritize and let you know anything you could potentially miss. 'Cause That could be an issue. I feel like it has a, from a, what would you call it, a tech experience level, it's got a wide appeal across tech experience levels. And I do appreciate the earcons or the boops and clicks and stuff like that. It's nice to get that feedback of, yes, it's doing something. It is thinking. Your email did in fact send. I've I really appreciate that you all added those in.

Sriram Ganesh: Yeah, thank you so much. The other thing that I also wanted to call out is that the product is only as good as the input that we get from users, right? Like one of the— and so within that, one of the innovations that we did, which really pays us a lot of dividends over time, is how users can give feedback to Lumin, right? And if you look at most products, the reason why they suffer is because developers tend to add a feedback button somewhere in some settings page under some button, and it's up to you to go find it and actually draft out feedback. For a person with disabilities, it's even harder to go find that. Generally speaking, you want to give feedback when something's not really working well.

None of us are going, hey guys, you guys did a fantastic job, even though we get several feedbacks like that. I think the ones that we are most interested in is when did Lumin let you down? And we want to know that immediately so that we can actually go and fix it so that other users don't face that. And so what we did was we actually made it part of Lumin. So the way to give feedback to us is you just tell Lumin you want to give feedback and Lumin will collect it and send it to us. And, and so as you're talking to Lumin, if you don't like something, you just go, Lumin, I don't like the way you did that. I want to send feedback. Lumin will be like, oops, Let me collect the feedback. It'll collect it. It'll send it to us. And within a day, you should actually receive a response from me telling you that we're working on it, right? And that gives users more confidence, which gives us more usage and more data. And this is not data about like your email. This is data about like Lumin messed up with this. So go fix that. And it's the reason why we are able to improve the quality of the product so much. It's because of this feedback loop with the community.

Tony Stephens: Wow. On the issue of privacy, where are there hesitations? You know, that's a big safeguard when we're opening up our personal lives, which email can be extremely personal or confidential for work and things like that. How does it work for safeguards with privacy and things like that?

Sriram Ganesh: We've thought about this very deeply. Firstly, all of our work accounts are connected to Lumin. We send really critical stuff to everyone. We are telling friends and family to come on to Lumin. And so we take this responsibility very seriously. And we've thought about it in a few ways, right? Like, I think from a privacy standpoint, the first thing that we do, the reason why you always hear the boops with Lumin is Lumin is actually going and fetching your emails from the Google servers, which means that we don't actually move your emails anywhere. We securely access it from Gmail servers where your emails are already sitting. And we use it for when we need it. And then when we don't need it, it stays in the Google servers. We don't store it. We don't keep it. We don't really know what's happening with your emails unless you specifically ask for it. We don't go around touching that at all. The second part is that what do we actually save? What we save, we don't actually save a lot of information about you. We only save preferences and rules that you would like. So if you say, hey, Lohman, I don't like emails that are so verbose. I like emails that are shorter. I want my signature to have my company email ID or my company domain name added. I want my voice speed to be 1.2x. I want my voice to be, say, voice option number 3. So these are almost preferences where that we are constantly asking you and storing.

So the second part of it is, how do you know what we're storing about you? Super easy, just ask Lumin. Lumin will tell you everything it knows about you, just verbatim. And you can go, and if you don't like something, you can just say, hey Lumin, I don't like that, forget about that thing. And it'll forget it. And the way you can always check is you ask it. And finally, we care about control, which is that for some reason, if a user is like, you know what, I tried Lumin. It was fine. I don't really want to use it anymore. But I don't really like that you guys have like an access to my email that you could use. Please feel free, go into the settings button. There's a big button that says delete. If you delete it within 24 hours, we wipe everything off our servers. And so the way we look at privacy and security is that security is utmost important. And so we are keeping your data locked in Gmail's servers where it's already sitting. Transparency is you knowing what we are saving and why are we saving it. And we keep telling users the only reason why we're storing something is because we want to make your experience better. It's not about sharing the experience with other people. And finally is control, which is for some reason, if you don't like things, it's a one button that you press and we wipe everything out. We also have, you know, users are sometimes worried about like, hey, with this AI thing, like, is my data just going to like some big model and are they just stealing everything and training on it? Is that happening? I don't blame users for asking these questions because I think generally speaking, Silicon Valley and the AI industry has not done a really good job at talking about AI safety and privacy. And so people should ask these questions.

From our standpoint, We try to be very proactive about what we're doing, why we're doing it, and how can you change, how can you be in control so that you're not regretting it. We actually have contractual obligations with all of our providers where even the frontier models that we sometimes use, where they cannot train on our data and we pay them more money in order to get this privilege on behalf of our customers. So that's a long answer for how we think about privacy and what we do under the hood.

Tony Stephens: No, that's important. I mean, and safeguards too, just for folks, because there's been a lot of stuff in the news over the past month with like the Hugging Face and ChatGPT, like Lumin, there's no risk of Lumin going rogue or anything like that. I would imagine like what safeguards are there?

Sriram Ganesh: It's one of the reasons why like we've just not, it's one of the reasons why Lumin does not have access to the internet. Lumin has access to one door, which is I can go into— when Tony asks for it, I have permission to go into Tony's email and fetch him exactly what he wants and show it to him. If Tony wants to take action, I'm going to take that action and be done with it. So Lumin is not unlike some of these other products. Lumin is not loitering around your data when you don't ask for it. So if you don't use Lumin for the next 10 days, Lumin's not touching anything unless you call up and say, I need this.

Tony Stephens: Yeah. Wow.

Tony Stephens: Very cool. Just in the sense of, I keep thinking back to the demo you did and its ability to just digest information. And I mean, I think I'm going to be sending myself PDF files. Just, I emailed myself a lot of stuff on weekends that I will take, you know, it's kind of like an email take-home to work from home kind of stuff, you know? And now I'm just like, oh, I'm just gonna bust this open and be like, I don't know, it just, the seamlessness, right? That I can sit in the comfort of my chair, not be wrestling with going into like Chrome or something and trying to find things in my accounts and all that stuff. And can even get down to like reading documents, summarizing documents, describing photos and things.

Sriram Ganesh: Like, I always love talking about, you know, like reading a PDF for me is still like a feature, but I love talking about how these features impact our audience, our users' lives, right? Like, you know, just a few days ago, one of our users emailed us their story and they told us how because of their eye condition, it became really hard for them to get a lot of work done in their home because contractors would send you something and you need to like review it. You need to make sure you're not getting charged too much. You need to negotiate. You need to be able to look at different things. And so he had a new project in his home where he was getting his his roof replaced. And he got quotations from 6 or 7 different vendors with different prices and everything. And that he would need help from his kids or someone else to kind of actually end up doing that. And in this case, he was just, he got everything by Lumin. And he was just on top of everything. And he was able to get it done, feel good about it, which is very important to us. Like we care more, we also We care deeply about how Lumin makes you feel in terms of feeling confident, feeling independent apart from doing the job because we all, you know, like our condition should not should not restrict us from doing the things that we have. And like when you are able to do those things, you feel useful in your life. You feel useful in your to be in your family's life. And like it. And so for me, it was just like that's a fantastic story. If Lumin was able to bridge the gap between you and your contractors by getting to read all the quotations. looking by it line by line, finding discrepancies so that you can go and bargain or negotiate your way. That is fantastic.

Tony Stephens: Yeah. Where do you envision this growing in terms of functionality? Are there other platforms? Like, you know, a lot of people use other things. Like we use a service called, or a service, not a service, we call it a platform called Trello for like, you know, where people are also posting documents and doing links. Do you feel that you have focus in perfecting this space, or are there ideas or dreams of you, particularly as a user now with lived experience of vision loss, like where this could grow? Or, you know, in other spaces, other people could maybe generate stuff if this is your lane and you're going to stick and make this the best it possibly can be. Um, you know, where's there green grass out there, I guess, for, for new exploration and how this could be even further use of technology and AI for our lives?

Sriram Ganesh: I think this is a great question and something that we internally discuss a lot. You know, we don't— we want to be great at doing a few things on behalf of the user, right? And so as we are expanding the scope of what we can do with email and calendar, firstly, we are adding more services. So we are, in the next few weeks, we will have Microsoft Outlook, which is our next big upgrade, and Microsoft Calendar. And so users will be able to, you know, we've got a pretty big waitlist of users on Microsoft who are very excited to try Lumin. And so one part of it is just increasing the number of providers. The other part of it is going in depth into email, adding features like the ability to be able to snooze stuff and things like that, right? Like unsubscribe, snooze. You should see a lot of that coming up in the future. But what really excites us is to look at opportunities where the email and calendar product can become the platform, right? Because email and calendar is not the product in many ways. We think about what are these big jobs that we can do on behalf of users. A few of them that we are broadly looking at is, one of the big problems that we see users constantly facing is filling forms. The Quagmire report talks about this in depth in so many ways, whether you think about recruitment, which is obviously a huge challenge. Employment is a huge challenge within our community, and a large part of that is you know, the process of recruitment, it is so inaccessible. It mostly involves going to websites and forms and doing all of these things before you get to talk to a real human.

The second one is really around healthcare, right? Like people with disabilities need more help. It's like they're the ones who need more help with healthcare, but their experience with healthcare is completely inaccessible, including getting benefits, including making claims. And so, you know, our dream is to actually go for those bigger jobs. It's like, you know, email and calendar will continue to be the base of the product. But, you know, what are these really big problems that— so if you really think about it, what Lumin is, Lumin is an easy way for you to access complex information. That's what it is, right? It's like, just say something and we'll go hunt for it and we'll bring it back. What our internal thought process is, how do we leverage this superpower to go do jobs that are much more meaningful for you, where the impact and the consequence is much higher. And so those are, so in the near term, you'll see us expand the scope of email and calendar. And our ambition in the midterm is to see how can we leverage email and calendar to do jobs on behalf of users, whether it's in recruitment, whether it's in healthcare, to really do things allow them to do those complex jobs, whether it's care management, whether it's insurance, on behalf of our users, keeping the same interface. So you should just talk to Lumin and be able to fill up a form in the future.

Aaron Preece: Something I'm curious about on like the functional side of things, say if I— because you mentioned like, oh, if you didn't use it for say like 10 days, it wouldn't, it wouldn't be doing anything on its own. So say I'm like away from work, like on vacation for a week, and I get back and I need to go through a week's worth of email. Can I ask it to go— would I need to ask it to go day by day potentially? Like, how far back can it reach? Or is that something it can like take some time, sort through a bunch of, you know, 100 emails or something and come back with the most important emails? Like, basically, what's the extent of like how deep can it go, I guess?

Sriram Ganesh: That's a great question. When you ask for your emails, we We can, so you can find your emails as far as you have emails. So that's number one. So you can absolutely find any email. But in your situation, when you come back after a week and maybe you have a week's worth of unread emails, what we do is you have to do it day by day today with Lumin. But within a day, we prioritize all the important stuff. And we have an internal kind of like logic for that, like emails from people stay on top and ideally promotions and junk stay in the bottom and then everything else in the middle based on priority. And so when you ask for emails from today, it should give you the top emails from today and then you can just clear out, you can just batch archive is what we do towards the end. We're just like, hey, these 2 batches mostly junk, clear the whole thing out. It's out of your inbox and then you go to day 2, it pulls up the most important thing and we'll keep expanding this and keep making it better over time. But yeah, so today you can ask for it all and it'll give you the prioritized emails on top and then you can just clear out the stuff in the bottom.

Aaron Preece: So that makes sense. So then you just keep going back day by day today, then go back to yesterday and then go back to the day before that and just keep going through.

Sriram Ganesh: You just keep— and the way I like to do it is like, you know, I like Lumin leading me. And so I just go, hey, Lumin, tell me what's new in my inbox. And then Lumin's like, here's 5. Do you want to do this? I'm like, yeah. And then it says, do you want to go back? I'm like, yeah. So it's mostly just me going, yeah, no, skip, send a reply, archive. And that is also by design because while audio is an incredibly accessible format, it also takes a lot of energy to keep talking, right? So if I don't need you to continuously keep talking, I shouldn't demand for that, which is also why we allow for some like small commands. We also teach users to say things like, hey, just say what's new. What's new is like this. It's a command. It's a small little command. But when you say what's new, we pull it all, prioritize it, and give it to you in a nice little beautiful playlist of emails because it's an audio environment. So it's—

Aaron Preece: I imagine that helps with latency too, because I know sometimes like if I ask, say, ChatGPT to read like 6 documents and give me some details. It will take a couple of minutes. It seems like to keep that conversational good latency, you'll need to break it up into more segments for that reason too, I'd imagine.

Sriram Ganesh: Absolutely. It's about— again, it's the other harder part of our job. With audio. It's why Lumin first gives you a summary of something because like if every single email, the only way for you to know what is happening is either reading the subject line or reading the whole email, like both of these are not viable options because sometimes subject lines don't really mention what this email is about other than just a subject line. And reading the whole email sometimes is kind of pointless. And so Lumin tries to give you a summary. But the important thing is that the summary has to be accurate. Because if I'm replying based on the summary, then, you know, it's all of those trade-offs, which is we want to give you gists of your email, summaries of your email. You can, of course, say, hey, just give me the whole thing as is and it'll read it out to you. But when you start like living in the audio environment, you want to just because you also want the advantage of speed, right? Because audio is inherently slower compared to the visual medium.

Tony Stephens: Yeah.

Sriram Ganesh: You can see many things together. But that's the other reason why we started with our audience in specific, because we're not interested in being kind of an okay solution for someone with vision. We would rather be a really important solution from someone with— for someone without too much vision. Because that is, even though that's a smaller audience, that audience needs us and then we can have a great relationship and build something meaningful.

Tony Stephens: Yeah. Wow.

Tony Stephens: Well, again, excellent work. I mean, and then sharing it with us and really walking through the product and letting listeners know more about it. Where can folks go to learn more about Lumin?

Sriram Ganesh: Lumin is currently available on the Apple App Store. We work on iPhones. We support Gmail and Google Workspace accounts and calendar with support for Microsoft Outlook and Outlook Calendar coming very soon. You can find Lumin on our website, trylumin.ai. That is T-R-Y-L-U-M-I-N dot A-I. And you can find the App Store link there. And in case you're an Outlook user, you'll also find a waitlist there. Just add your email in there and we'll reach out to you as soon as we're ready. Lumin is currently free for anyone who's coming on board as part of early access. And so, yeah, please come on board and try it out.

Tony Stephens: Excellent. Well, we'll have links in our show notes for that information as well. People can try it out. And again, Sriram, so nice to meet you and get to know you and your story and really just what you've created in the sense of, again, the user experience that's really tackling so much in a way that AI is on that positive benefit side amidst all the doom and gloom in the world about AI. It's refreshing just to hear something that's so clean-cut and to the point and is really user-friendly. So thanks and congrats on just really, really giving the community a product that I, my feeling is for me, I already see use for it. So that's great.

Sriram Ganesh: I love it. Thank you so much for, thank you so much. Thank you for the great conversation, but also thank you so much for the love and support. We really appreciate it. It's, it's at the end of the day, that's why we're doing it, right? It's like we want, we want our community to kind of like what we're doing because that's our job. And it's, it's great to see at least, uh, you know, we have a long way to go, but, uh, it's been a good start.

Tony Stephens: Yeah, I mean, it's exciting to think 6 months, but where are we going to be in 18 months and 36 months? And yeah, so, so we'll look forward to hearing back on what, what other stuff is in your, your space and you're, you're pushing out for the world to experience. So Thanks. And again, folks can check that out in the show notes for our podcast. Be sure to like and subscribe to AccessWorld wherever you get your podcasts from. AccessWorld is a production of AFB Studios at the American Foundation for the Blind. And you can check out, like we were talking at the beginning of the podcast, the latest issue of AccessWorld magazine, our quarterly magazine on digital inclusion and accessibility, at afb.org/aw. But for now, as always, Aaron, Thank you so much, Sriram. Thanks for joining us. And everybody, we'll talk to you in the next episode in a couple weeks. Take care.

Outro: Thanks for listening to AccessWorld, a podcast on digital inclusion and accessibility, a production of AFB Studios at the American Foundation for the Blind. AccessWorld is edited at the Pickle Factory in Baltimore, Maryland, with digital media support from Kelly Gasque and Breanna Kerr. Questions or comments? Email Or email us at communications@afb.org. AFB.