A Blind Founder Builds An AI Assistant You Can Trust

Your inbox is where productivity goes to die, especially when the tools assume you can constantly scan a screen, hunt for tiny buttons, and triple-check every draft. We sit down with Sriram, founder of Lumin, to unpack what happens when an AI voice assistant is built for blind and low vision users first and held to a higher standard: it has to be accurate enough to trust, fast enough to use daily, and deep enough to handle real email workflows instead of shallow commands. 

Sriram shares his journey from video game engineering and immersive tech to entrepreneurship, then the turning point of progressive vision loss from cone rod dystrophy. That lived experience shapes Lumin’s core idea: “talk to your email” in a way that actually gets work done. You’ll hear a live demo where Lumin summarizes new mail, opens a message, describes a photo attachment clearly for social media, drafts a reply, takes an edit, and sends only after confirmation. The conversation digs into why this kind of AI productivity tool is fundamentally different from generic chat apps and why handling attachments, revisions, and back-and-forth is where most assistants break down. 

We also get concrete about accessibility and inclusive design: co-design with Lighthouse for the Blind, screen reader support, robust text zoom, and a simplified interface that keeps power in voice commands. Then we look ahead at the product roadmap, including Outlook and calendar support, plus the longer-term vision of meeting users on more devices like Android, smart glasses, and eventually cars. Finally, we tackle AI trust and privacy head-on: clear disclosure of cloud providers, visibility into what Lumin remembers, and user control including account deletion. 

Subscribe, share this with someone who lives in their inbox, and leave a review if you want more honest conversations about AI, accessibility, and assistive technology that actually works.

Transcript

Neil Milliken (00:00)
Hello and welcome to AXSChat. I've got a super guest with us today, Sriram, who is going to tell us about his journey and the the product that he's created Lumin. I I'm really excited about this. I had a chat with Sriram a few weeks back now, and I got to have a play around with with Lumin. it's one of our favorite topics at the moment, it's AI-based, so you're gonna get a good flavor of AI.

and assistive tech. But Sriram, welcome to AXSChat. Tell us a little bit about your background and and your journey and and and what taught you know what what caused you to sort of come into the space of creating technology, particularly focused on on inclusion.

Sriram (00:46)
No, absolutely. Firstly, thank you for having me on this podcast. It's great meeting you guys. I'm Sriram. my my journey started as as a young kid in India. I've I s my my dad bought home my the first computer that I ever saw when I was 13 years old. I saw Sesame Street on it, and even though I was a little too old for that game, I was instantly hooked. watching Elmo jump around on the screen was absolutely fascinating for a 13-year-old

kid.

And that's what led just like many other computer engineers, that's what led to my fascination in computers. And one thing led to another. And I was and I was so excited about computers that I wanted to learn how to make games. I wanted to be one of those guys. and I studied video games engineering, built video games for a while, spent a bunch of time in Scotland in Dundee.

Building video games. When I was 20 years old, I showcased a video game that we built called Color Scheme at the Edinburgh Fringe Festival, which was fantastic. More than 10,000 people came and played our game. It was huge for a for like a 19-year-old student. That led to me coming back to India, starting my entrepreneurial journey. I've always been into kind of visual computing, building things like video games, augmented, and virtual reality.

Try to stay at the cutting edge of technology just because I'm so fascinated. I've always kind of been interested in building products and technologies that were almost magical. You know, the person on the other side should feel like it's something really exciting from the future. That's kind of what I wanted to build. That my previous company, Who That, which was in the augmented and augmented and virtual reality space in my previous decade, grew and then eventually successfully got acquired by a pretty large education company.

Where then I got to spend four years building education games for kids, which was my two favorite passions. One was video games, and second was how much I kind of disliked traditional education. And I really wanted to build sort of interactive education to kids. my parallel story of is that I started losing my vision a few years ago. completely by surprise, it was turns out when I turned 30 years old, the the genetic condition that I've always had started to mature.

it's a condition called cone rod dystrophy. It affects my central vision progressively. And that made using computers incredibly hard. And over the last few years I've kind of been tinkering on the idea of doing something to help this community while also being useful. and we started Lumin. That's my really long introduction about me.

Neil Milliken (03:08)
No, but it's it's great to ha it's great to have the background. and you know, and the fact that you're a s successful entrepreneur in the first place makes it better for you starting up in in the assistive tech space because one of the things that I've seen over the last couple of decades is that a lot of people come into the assistive tech space without the business background. and that means that they they really struggle to get traction

or they

They have a great idea but they don't know how to implement it. They don't know where to get funding. They don't know how to sort of do all of those other things that contribute to the the the success. And we we really want, you know, people that are innovating in the disability space to be successful. I I know Deborah's got a question and and and then perhaps we can go on to actually what Lumin is. But Deborah, you first.

Debra Ruh (03:57)
Well, you know what? Actually, that is the question I had. Now and maybe we'll do it backwards because I Neil had told me about it and that you're gonna be a guest. So I went out and looked and I was like, well, before I went and looked, I had already made a decision. I didn't understand the product because can't I talk to Chat GPT and Gemini and perplexity and Claude?

So it's like, well, okay, that's great. And I see AI being so beneficial, starting to be so beneficial for our community. And so I was I was really intrigued. So actually before the show, I went and asked ChatGPT and ChatGPT explained to me a little bit about how wrong I was. And so I'm sort of asking out of order, but tell us about what it is and why it isn't what I was thinking and something that's available everywhere.

Sriram (04:48)
This is a great question and something that we we we we hear a lot and you know and and the easiest way to think about it is that you know we've been using these voice assistants like chat GPT and before that was Siri and Alexa for a while. And for most people, if you think about it, you've used Siri and Alexa to do the same few things again and again. You set a timer, you set a countdown, if you're cooking something, you want a timer.

If you want to be reminded about something, you do reminders. And so these are great to set up either you just want to have just a conversation about really nothing in specific, or you just wanna take a very quick action. Lumin came out of my own personal necessity of just having to do email properly. And what that means is firstly, you have to get my emails right all the time. You have to connect to my emails. Second, you have to give me a high level summary really quickly.

third, you have to go in depth, give me more details. Fourth, if there's an attachment, read the attachment. you know, if there are multiple people on the email and I want to switch them around, let me do it. And so we really think about it as we're building a voice assistant that can help you get work done, starting with email. And work is a lot more depth than shallow. So I'd love to show a quick demo. That's the easiest way for anyone to get a sense of what Lumin is.

Debra Ruh (06:09)
That'd be great.

Sriram (06:11)
Yeah. I'm just gonna take maybe less than a minute and just show you guys a really quick demo. This is all happening on the Lumin app that's available on the App Store. So none of this is fabricated. It's connected to my actual inbox. and so, you know, this is as real as the demo gets. let's go.

Good morning, Sriram Hi Lumin. what's new in my inbox?

Seven new emails in your work in box since yesterday evening. One, Daniel Semro wants to be unsubscribed from your emails. Two, Pat Quinn sent a photo from NFB and thinks it could work for social. Three, Guy Castingway is asking whether Lumin will support calendar in the Let's see what he's saying.

That says, Morning Sriram, check out this photo of you at the NFB booth. They think it might be a good candidate for social media. There's a photo attached. Want me to summarize the image?

Looks like a candid conference booth shop. You're seated behind a black table wearing a black lumen top, sunglasses, and a lanyard, and you're talking with a woman standing in front of the booth in a pink shirt and a long floral skirt. On the table, there's a laptop, a QR sign that says talk to your email, some snacks, and coffee cups. The background shows a busy event hall with other booths and a sign about smart glasses. this actually almost even audibly sounds like a great shot. And yeah, just reply to Pat.

Let him know that this looks pretty good and yeah, we should get it up and going.

This is going to Pat Quinn. I wrote, Looks pretty good. Yeah, we should get it up and going. Ready to send? Pause this a little bit.

This is going to Pat Quinn. I changed it to. This looks really good. I think we should post it. Ready to send? Yep.

That's said. Back to your inbox. You know what? I'm in the middle of a podcast, so we need to end the call right now. Can you end the call, please?

Debra Ruh (07:44)
Yeah.

Sriram (07:46)
Okay, talk soon, Sriram. That's a quick demo of Lumin.

Debra Ruh (07:47)
Okay.

Okay. So I I I'm going to say just two quick things because I promised in the background, now I'm going to turn it over to Antonio or Neil. But I actually use an AI tool. I won't name the brand right now. And it is it has been wonderful. It has saved me so much time, maybe about seven to ten hours a week. But after that demo, please, please may I join up? Where do I go? Wow. Wow.

Sriram (08:10)
Ha ha ha.

Debra Ruh (08:12)
Wow. Wow. I am so impressed.

I'm so impressed. That changes the game. So all right, I'm stopping, going over to you, Antonio, I believe, right?

Antonio Vieira Santos (08:20)
No, yeah my question goes somehow in the same path than you were thinking, Deborah, is today we have we are flooded with tools and systems that are that tell us that we are able to improve your productivity. So in that case, what my question to you Shridan is what are you bringing new and useful that distinguishes yourself from the rest?

From the crowd, I would say.

Sriram (08:49)
We are the most reliable way to do email by voice today. We support many, many types of attachments. You can you can say read out an attachment in page number four. We support if you have a if you have Microsoft slides attached, we'll read page four of the Microsoft slide. So you can actually get slides read out to you. You can get Excel sheets that are attached read out to you. And for a person who is visually impaired, the what you don't need is a shallow experience. You need your work to

be done.

If there is an attachment, get the attachment done. if there is an Excel sheet, get that done. make sure you never ever mess up my email. As someone losing his vision, I need the product to be so re so reliable that I don't have to go and check with my eyes if it got it right. And with most of these tools that are trying to do a hundred things and are trying to get you up to be productive, we are trying to do

one thing at a time and we're trying to do incredibly well. I think that is where the difference is and that's why our users use our product today.

Antonio Vieira Santos (09:49)
During the process of developing your solution, okay, what steps? You know, you you have to live with experience, but you know, you are probably what they would I would say an expert user, a very savvy user. How you

Sriram (10:03)
Yes. Yes.

Antonio Vieira Santos (10:05)
work in a way to make sure that even someone that is not as savvy as you are with technology, they can also easily navigate and using use the tool.

Sriram (10:15)
Absolutely. We actually s even before we started building the product, we we we built a collaboration with an organization called the Lighthouse for the Blind, based out of San Francisco. We recruited 50 users from different walks of life who are going through visual impairment. and we actually sat down and co-designed the product with them. there's a reason why the product has only one giant call button in the app, is because we want to make it so easy.

To get into the product and to get out. and we instead of adding, say for example, we make design decisions like instead of adding a lot more buttons on the product, how can we make even those commands voice accessible? we also do things like we go through extreme lengths to support you know screen reader compatibility, zoom level of text.

much more than what traditional apps, if you go into Chat GPT and you increase your font size to maximum and you take a look at the app, it doesn't look very well designed. Because when you're trying to serve a billion users, the 10, 20, 50 million users who need that kind of interface are not necessarily your priority. But that user base is our priority. And so along the way, and a a simple example of that is just being able to end the call by saying end the call.

Because for most people that would seem like a very trivial feature. But for someone who's visually impaired to find your phone, find the button, scroll to it, tap on it, is a lot of pain. And when a lot of these decisions compound over time, you have hundreds of design decisions made with the community. So today, for example, we have users from lots of demographics giving us feedback in the early days. and that's how we make the product incredibly robust to the audience.

Neil Milliken (11:56)
and and Antonio, I've tried it. So and and and one of the reasons I was keen to get Shiram on was because I tried it and the onboarding experience was really easy. I've worked with speech recognition systems for twenty five plus years now. And you know what Shri Ram was describing at the beginning is like speech command, voice command. This is much more conversational, it's much more intuitive. you know, and and and you know, you really can

Be eyes free, right? I'm a visual person, but I can be eyes free with this and find my way around. And as an ADHD, that's you know, that's impressive because generally speaking, neurodivergent people are able to hold less stuff in their working memory. The fact that that Lumin still works for me and I can you know I can interact with it means that the the conversational design has been done really well.

know. challenge is for me that most of my stuff's on on a you know, a different email provider than the one that is currently supported by Lumin and and so I'm only really interacting with junk mail. but but but I you know it's I I was really s seriously impressed by how well it handles natural language and how intuitively the responses are

that enable you to then deal with the sort of the the aspects of work and handling workflows and so on. Right? I I I get, you know, I I like Deborah was equally skeptical when, you know, like when I when I first heard because likewise, like long term speech recognition user, long term user of AI tools. I was like, well what's the difference? And I think that you know the difference is in

just how polished it is, right? That's that's what you were saying, right, Sriram? It's the you you you're doing one thing really well.

Sriram (13:52)
Yeah.

Yeah. It it has to be so reliable that I don't have to keep second guessing. and with a lot of products that connect to a hundred of your services, I would actually like I would in many ways I would turn it back and say, what do you use Chat GPT for beyond just talking and searching? For in voice. of course on text you can do many, many amazing things. But you know, if if you were to almost think about like, you know.

With any of these voice tools. Sure, with City, you can you can send a message or you can send a quick command. But if you draft an email even with Apple Intelligence today, and if you want to go back and forth, change things, add an attachment, that's how real work is. It's not as simple as just giving a command and going things away. Things go back and forth. That's how you would work with a person, right? You would you would sometimes ask someone to do something, they would come back with something else. You would go back and forth, get it polished, get it fixed, and then send it out.

Things are rarely one shot. And Lumin lets you do that. It lets you go back and forth, make changes. You can brainstorm with it. You can put things back. But it's not just for chatting. It's it actually gets things done in the real world. So it can send email on your behalf and make sure that the spellings are correct and make sure it checks with you. Those are all the design decisions that go behind making it so accessible that a blind person can trust it.

Debra Ruh (15:12)
Well, that that's just so impressive. And I actually am a power user of voice when it comes to a lot of chat and Claude and others. and and sometimes I get way ahead of the AI and I confuse it. I know I confuse my Alexa often and because I just want to go on and on and on and but

I really was impressed with what you're saying. Like I think it's so powerful what you're saying that even a blind person can trust it. because so many of us, you know, I I, as Neil's saying, I'm I have near I'm very neurodiverse. And as I'm getting older, my ADHD is getting even more interesting for me to manage. But one and also as an entrepreneur and a business owner and a founder of a nonprofit.

With multiple emails and very complicated emails, that's very, very easy. I I was really afraid to trust my email to anyone. And I have trusted to a provider that's really, really helped. But what you're talking about is just sounds like heaven to me, a person that's neurodiverse. And and and I think how powerful this is for our community. And I also want to say one more thing.

and then let you respond because we're so excited we're forgetting to let you talk as much. But and I know Antonio has a question again, but but I think that what's so fascinating about it is what you're talking about would benefit everyone and how wonderful at the same time though, it benefits the disability community first. We're immediately included. And then I would also say to the disability community, these are the entrepreneurs that we support, we support our own.

We need to do a much better job of making sure we are supporting the entrepreneurs that are people with disabilities bringing us amazing amazing products. So I applaud that. So let me turn it over to you and then I'll let you turn it over to Antonio.

Sriram (17:02)
Yeah, I I think we've been very clear and you know, my experience has been in building products for consumers, for users. I'm I'm I you know, before like my my main gig before this was working with with kids and building educational products for them and they have their own set of unique needs, right? Like trying to teach a four-year-old kid. a four-year-old kid has no motivation to sit in one place. They're incredibly energetic and they want to run around and do things. But and if you wanna

teach them something without forcing education down them. You want to make it inspiring for them and you wanna make build something that is exciting and accessible for them. And so so much of my learning for building for kids who are in their pre-kindergarten, kindergarten, some of those design philosophies carry so well here because the user has different needs. And we need to meet the user where they are rather than the user having to meet the technology where it is. The technology is here to serve the users.

and we are very clear about that. And at the same time, we're also very clear that we don't want to just give you technology that is okay for you. we're like I live in a city where you know, there are Waymo, there are these cars that drive themselves. And I think that's the kind of future that inspires me. so many people from the blind and low-vision community are positively impacted.

buy those vehicles because of safety concerns and other things, right? It's it's fantastic. Like they get to sit in a car and just get dropped off and you don't have to worry about you know someone doing something. similarly, our goal has always been the the bar that we're trying to set is it has to eventually be so good that even sighted people may want to use it. and and the and in order for it to be that good, we have to make it so reliable within our

Community, but also keep the goal constantly as we're building consumer technology. We're building technologies that eventually people say, for example, if you're driving, we we call say say we actually are deep believers that there are many moments where people are situationally or selectively disabled. If you're driving a car, you probably should not be staring at a screen and tapping on it. But at the same time, what if you could just

Keep talking to your assistant and get work done along the way. So we have commuters using our product. We have some people who drive who use our product. So we can almost naturally see that if we design keeping accessibility in mind, we can actually design a universally great product. and we are trying to see how to achieve that. but yeah, that's what we're doing. Antonio would love to hear your thoughts.

Antonio Vieira Santos (19:29)
No,

so I based on what you were just saying, I I'm very curious to see to and to understand what's your what's your roadmap and what you have planned further as you develop. You you were you were talking about cars, so I I was actually wondering, well, maybe you will embed this in the operating system of the car, so it will be more effective for the driver to to use it than to to have a mobile phone. So I'm really curious to know about what's your product roadmap.

and to understand how how you plan to develop and what ideas you have for for the next version of your your system.

Sriram (20:03)
Absolutely. So we think of ourselves as as a software layer for voice. So we will meet users in different devices. Today we are on a phone, just because the iPhone is one of the most commonly available devices around and it's a it's it's easy to access. So I don't have to push you to buy another device or you know do any of that. So just download an app and you use it. But tomorrow

you could imagine us being on. So there are there are two aspects of it. One is how the software expands. So from a software expansion standpoint, you should you should assume that we will add more services. So I'd love to get Outlook in pretty pretty soon. lots of Outlook users. I'd love to get calendar in very soon and really expand the surface from there. depending on what sort of services users pull us into. I think the calendar and the email are

Fairly obvious and they work very nicely together. It's almost like a nice little assistant sitting with you who can help plan your day, plan your week, get emails done. and eventually we'll sort of expand to other services depending on what users care about. and from a device standpoint, today we're sitting on the iPhone. you can foresee us expanding to you know other platforms like Android, but also maybe to smart glasses.

Maybe into the car in the future. so all of those are expansion opportunities. But as a startup, just because we want to stay incredibly focused, the easiest thing for us to do is again expand too fast, doing many things and then not doing anything really well. And so there is almost like a level of discipline internally where we are like, we will do few things in the beginning, but we'll do them exceptionally well, and we will slowly expand the surface area.

And make sure that we add value. So you should see us expanding to more services like Outlook, Calendar, et cetera. So

Neil Milliken (21:46)
I think I think that's key. I mean the the the the the thing is the the laser focus on doing stuff really well. We a lot of software, a lot of services have a huge amount of bloat and cruft and you don't you know, you've got everything trying to be the the everything tools and so people are just throwing features at things all of the time. a and yet people still end up buying specific tools. And it's why this whole app economy grew because

The original software providers were doing everything but really badly. And I and so that created the opportunity for specific apps. And I think that that that opportunity to have focus and do it really well is is is still a like a really big opportunity. Going at a slightly different angle, and and you'd already talked about, you know, creating something that's so good and so accurate that

That users with low and no vision can trust it. You're using a lot of AI, right? And a lot of people don't trust the people that own the companies that are producing the AI tools. So we had a conversation offline about trust. But can you tell our audience about how you approach AI and which tools you're using and how you build that trust and keep that trust for your users?

Sriram (23:04)
Yep, very happy to talk about it. I'll talk and I'll I'll really talk about it in terms of like trust, tr transparency and control, right? It's it's ultimately

Neil Milliken (23:11)
Yeah.

Sriram (23:12)
all of this is about you as the user knowing exactly what you're signing up for and having the controls in your hand to say, I either opt in or opt out. That that's that's precisely what this is. and to to begin with, first we

you know, we don't dilly dally around the fact that we absolutely use cloud providers because that is the way that we can actually add value to our users. Adding value to your users is number one. we could absolutely focus too much on privacy and bring try to do everything on device today, but that would be the wrong decision because we would end up building an incredibly private private product that does almost nothing, which is which is an example of what City is to some extent today, which we want to avoid doing.

Instead, what we do is we use a certain set of cloud providers that sit between us and the user. an example of that is Google. but with each of A, we list out all our cloud providers literally on the first page of the app. We don't send you into a terms and conditions and ask you to read it. We we with the moment you open it up, there's like a small little pop-up that tells you. and the reason why we tell you is that we tell you that hey, we we are using these cloud providers, but

We also have enterprise level contractual obligations with these cloud providers to make sure that they cannot see or train on your data. So we do the onus, we pay them more money from our side to ensure that. The second aspect of it is what Lumin knows and remembers about you. And Lumin remembers things about you in order to personalize and make your experience better.

For example, if Lumin realizes that Deborah likes her email concise, it's probably important for Lumin to know that and for Lumin to not sort of go around doing different things, but remember that as a preference. And how does Deborah know if that's the case? You just ask Lumin. Lumin, what do you know about me? And Lumin will literally tell you everything it knows about you as a simple list. You can tell it to either remember it or not remember it. And finally, when it comes to control, we

We we don't have we only have a few buttons in the app, but one of those buttons is delete the account. And when and we've made it easy to find, but not easy enough that you may make a mistake and tap it. But once you hit delete account, we wipe everything off our servers and everyone else within 24 hours. So we start by telling you who we talk to, we tell you what we save about you and why.

And we give you complete control over how you your data sits with us or not. and that's what we and we think that is the right strategy to build something while at the same time, while because we need to use the cloud providers because that's where frontier technology is to deliver this experience, being transparent and giving you control is is where how we will build trust.

Neil Milliken (26:00)
Yeah, that's I think v really valid point and and I think that that way of being transparent with your user base is is is admirable. It's not always the case with the 300 page terms and conditions in six point type that you frequently just click through and accept. So we've reached the end of our time. It's been really great chatting with you. Great to have a demo that's gonna work on

an audio podcast as well as a video one. you know, where can people find Lumin? Is it the App Store? Apple App Store right now?

Sriram (26:35)
The easiest way to find Lumin is go to our website, trilumen.ai. That's T R Y L U I N dot AI. once you reach our homepage, we've made it very easy. If you have an iPhone, just tap on the App Store link and get the app. If you are an Android and an Outlook user, we're accepting wait lists for that. So give us your email and we'll be in touch with you soon. That's it.

Neil Milliken (26:58)
Great. Thank you very much. And I need to thank our friends at Amazon for supporting us and keeping us on air. And hope that everyone that is listening to this if you enjoyed this content, please, you know, subscribe, follow us and and share so that the community can learn more about amazing things like Lumin and learn insights from our guests that we speak to every week. Thank you very much.

Sriram (27:18)
Thank you so much, guys.