Embedded - 55: Embedded Systems and Cricket

Episode Date: June 12, 2014

Radhika Thekkath, CEO of Agivox, joined Elecia to talk about her start up, entrepreneurship in Silicon Valley, and getting NSF SBIR grants.  Contact Radhika  SBIR home page

Welcome to Embedded, the show for people who love gadgets. Our guest this week is Radhika Teketh, CEO of Adjavox. I'm going to ask her about her startup and about getting an SBIR grant. But before we begin, I want to remind you listeners that on the last show, we announced that SparkFun is giving away a $75 gift card so you can buy a kit to make an RUOK widget yourself. The contest is to send me a name for your proposed system and a description of what you put it in, you know, the creature. So name and description. More info towards the end of the last show, episode 54 with Elizabeth Brenner. Get your answers to me by 6.13. The winner will be announced 6.18. And now, back to my guest. Hi, Veronica. Thank you for joining me.

Hi, Elle. I'm happy to be here. Thank you for inviting me. Could you tell us about yourself? Sure. I have a PhD in computer science, and I've spent most of my time in an embedded processor IP company, MIPS Technologies. This was the era when MIPS was spun out of SGI. So this is the post-SGI days, this was the embedded MIPS. And I was director of architecture there for a few years. And at some point, I decided to quit and do some stuff on my own. So that's what I've been doing for a few years now. And I'm really enjoying myself.

So you did some consulting, contracting sorts of work after MIPS? That's correct. So after I left MIPS, I did do some consulting work. And mostly in the embedded area, I also did some patent consulting, you know, patent litigation consulting around computer processor architecture type of topics. I want to ask all these questions about computer architecture, but I also want to know about your product. Agivox, am I saying that right? Agivox is correct. The agi is the root of words like agile.

So the intent with agi was move, movement to be agile, and vox's voice. So AgiWox was meant to be voice on the move. And what is our product? So our product is really a dynamic news radio that we create on the fly, dynamic, from a user's topic interests so so i have a news reader uh where i read as rss feeds correct but it's not voice it's just i pick up my computer tablet or whatever and and read whatever i want and i know what i read is different than what my husband reads and different than anybody else reads. Sure.

So does Agivox work kind of like that? Is it RSS voice or something else? So currently we are RSS news-based, but we bring two different aspects of news together. One is we allow someone to express their interest in certain topic areas. So you could say, I'm interested in cricket, and I'm also interested in processor architecture. And what we do is based on what you've said that you're interested in, we create a special, a personalized news radio stream for you. And this is based on RSS feeds. So because you expressed interest in cricket, we would be intermingling cricket sports RSS feeds with systems, processors, you know, maybe

something from WE Spectrum, RSS feeds together to create your own radio station. So the second aspect of what we do is we're actually converting this written text into audio. So into synthesized audio, which is why I call it radio, because it's not a reader in the sense you don't read it on your computer screen. But you can listen to it when you're driving, when you're gardening, when you're knitting, when you're cooking, I mean, at any point when your eyes are busy. So, exactly with all the people who are listening to us right now, I'm not sure I realized that I was inviting you on to replace us, but okay. No, no, no, I'm not replacing you. In fact, you would be a source for me,

right? If somebody was interested in embedded news, then you would be a source for me right if somebody was interested in embedded news then you would be a source i would take podcasts as well in fact we take podcasts now so people put podcasts on rss feeds and we would source that ours is and so it would go into yours and they would get the little short description and that would be read aloud to them? Yeah, I mean, based on the description, we would be able to match people's interest with your particular podcast. So if we sourced your podcast, then if someone said, tell me about embedded stuff, we would be able to automatically figure out,

based on our machine learning algorithm, that, hey, this is a podcast that covers this embedded topic. And we would slice your podcast into the rest of the channels that we're reading to this person. So the interesting part is this intelligence. I came to it thinking that the interesting part was the voice and the commands. No, no, no. The interesting part is the intelligence. So the ability we give someone to be able to say something like, tell me about cricket and tell me about embedded news. We then take those words and run it through our topic discovery

data and information that we've already have on our server and based on that we decide what rss channels we're going to pick out for this specific person for l wants to hear about cricket and embedded news so she would get a special radio station well i might come along and say what i really want to hear about is traveling in europe and gardening. So I would get my own radio station on the fly. And you could do it on the fly. Gardens of Europe. Exactly. Right. I'd be into that. So our system is built in the cloud. And it's really based on a machine learning algorithm that looks at tons of data, news and blogs and things like that.

And based on the data, it figures out the topics that this data belong to. So then when someone asks for their specific topics, we're able to do the matching and then do the reverse matching to the channels to create the dynamic stream. But we're also synthesizing the text to audio to create the radio station and we can use podcasts how is this better than searching for news on google you can do this when you're driving and so you can interact with it when you're driving you do voice commands as well as voice output where you exactly what what kind of commands do you support um basic asking for uh asking for you know there's basic control commands for instance you can ask it

to pause and play and things like that but the interesting voice-based communication is when you ask for some topic of your particular interest. So I could say, I want to hear about basketball. Oh, and so I don't have to set these up ahead of time. I can do this while I'm driving. You could do it while you're driving. You can set it up ahead of time. I mean, you could do it either way.

Well, you know, ahead of time, embedded systems and crafts and a number of other things would be interesting. But then you hear something like, oh, I want to know more about that. And so you support this drilling down into deeper things or changing topics mid-drive. Exactly. Well, that's kind of neat. Or if you're listening to regular radio, a radio station when you're driving, and, you know, typically a news station talks about many different things, they probably spend 10 to 20 seconds on a topic and they move on. And you say, application and then say, tell me more about Nigeria. And we would really source at that point various newspapers that have perhaps an Africa section, a New York Times Africa, or the Times London has an Africa section. Or maybe there is a newspaper in Nigeria that's already in our system.

So that's what you will then get. And so we get news from the source instead of the mashed together, watered down stuff that I can often get by searching on a topic in Google. Right. So the distinction here is we are not a search engine in that sense. I'm not providing you with an application that brings you a pointed response to a pointed question like Google or Bing or any of these regular internet search engines. What we're creating is a radio station around the topic that's of your interest. Kind of like Pandora for information.

Exactly. Pandora for news and blogs. And you're CEO. Is that right? I'm CEO and co-founder. And I have a co-founder, Kathy Faust. And the two of us have been working on this particular product for about a year and a half now. down this entrepreneurship route about four years ago, the original product that I had has morphed into this. The original product was, as you were thinking, just an RSS reader. It was just synthesizing RSS feeds into audio, but we've discovered, well, we discovered along the way that what makes it more compelling, more interesting to people is the ability to provide them a way to dynamically pick their topics and to pick a dynamic station. And you can still set it up ahead of time, but you can also do it dynamically.

So with the voice generation, it is voice generation. You're not having people read these stories, are you? No, we're not. It's synthesized. It's text-to-speech, you know, synthesis engine. It's not our synthesis engine. We use an off-the-shelf synthesis engine. Do you use different voices for different things, or is it all pretty much you choose a voice and then you go on? We don't allow, at this point, a user to pick a voice.

We have two voices. We have one male and one female voice that we alternate between oh so you can tell which blog is which or you can tell when things change we're not necessarily matching gender to gender but we try to that would be hard yeah i mean it's not that hard but we try to alternate so that there's a change in what you hear. That's another way to also say, oh, we've switched to a different news item now. I wonder if you're ever going to get to the point where, as part of the RSS feed, that I, as a provider, would say, oh, well, if you're going to read it aloud,

I want this voice to be me. We have these visual avatars on Twitter and whatnot. I want an audio avatar too. That would be neat, isn't it? So about a year and a half, what is your timeline for the future? You're not open yet. You're not for sale in the Android store, are you? We haven't really been focusing on our smartphone mobile application for the last six months six to eight months because we've been focusing on providing a subscription-based api service to companies so what we really want to do is our initial target market has been automotive companies

oh okay so we were trying to market to and have been marketing to automotive companies so that they can embed this directly into either their head units or play around with it. If they like it, then they would ultimately adopt it into their systems. That's such a slower market than the cell phones and the constant thrashing of apps. It is a slower market, but there are are 20 000 news apps in the app store right so to be heard above that noise is also difficult and it's not it's not as though we are not building apps we are building apps and we will have apps in the market sometime actually fairly soon because the apps are well along the way. But we wanted to make sure that the apps and our marketing to automotive companies and to enterprise through the enterprise market complement each other. You know, there is some pressure on automotive companies.

For example, they've almost universally brought Pandora into their systems. And Sirius Radio. They've brought Sirius, you know, so why do they do that? Because they have been pressurized by the user market that outside the car environment use Pandora, use Spotify, you know, use these iHeartRadio kind of things. And so our goal is, you know, we can have the two markets sort of play with each other

hear about us from each other in some sense oh yes i'm always really excited when people say they hear about my podcast because of an automotive app because some of them do have podcasts exactly cool yeah things like stitcher right yes it's stitcher primarily and so the plan for the future is automotive and apps or is there a bigger plan more so that's the first step you know as a startup we've realized that we have to focus otherwise we would die and ignoble death so is it just you and and your co-founder, Kathleen? Catherine? Kathy. Kathy. Kathy.

So at the moment, it is my co-founder and I, but we have had a few people work with us and leave. Things go back and forth depending on our funding situation. And we are actually hiring at the moment. We are hiring two people, a programmer as well as a senior backend server engineer. Programmer for apps, for automotive? What kind of programmer? Well, more a generalist because the specific programmer we are looking for needs to be able to go between someend server stuff that's in Python and some Java to be able to take the apps that would be available and fix some issues and provide support and things like that.

It's tough to find software engineers right now in the Valley. It is very tough to find software engineers. And you were self-funded for a while. Is that right? That's right. We have been self-funded for some time. We've also raised some money from an NSF SBIR grant that you mentioned earlier. So the SBIR grant, that's the small business something. Innovation and research. Innovation and research. And I was first introduced to them, gosh, I'm not quite sure first, but at ShotSpotter, we looked at them as an option because there are a lot of SBIR Department of Defense grants, and they do want to support small businesses, but they also want you to jump through a whole lot of hoops. You say NSF SBIR.

I know they come from different government organizations, agencies. There's the farm group. Those are always really fun to read. If you search for the Department of Agriculture's SBIR grants and you search for embedded, those are hilarious. But I didn't know, how did you get hooked up with the NSF ones? So, you're right. The SBIR grants have been given out by various government departments.

You know, the Defense, Agriculture, Health, Education, and NSF. And I might not have named all of them. No, I think there's some security ones too, but of course we can't talk about them. Right, there are more. The reason I looked at NSF specifically is I found that NSF SBIR is one of the most open calls for application. In the sense, as you pointed out, the SBIR calls from, say, the DOD or DARPA or the National Institute of Health or Department of Education are usually targeting very specific technologies every year.

As though they're already targeting something that kind of exists. They're hiring their friends or companies they're already working with. Not necessarily. I don't think it's that i think it's that they believe that they want to move technology forward in a specific area that year so it could be you know laser weaponry you know if it's DARPA or DOD or if it's education it could be learning via online methods and understanding how that works. That could be the thing for the day or the year, right? And NIH may be interested in some specific aspect of something. I cannot even speak to that.

So when I looked through all this, I found that NSF, as the National Science Foundation, had one of the most open calls in the sense that if you had a research project that you could propose, you could pretty much fit that into one of their categories. And open research is what NSF encourages, and NSF seemed like the best place for us to target what we wanted to do. Your product seems awfully practical for science research. Do they kind of straddle engineering and science? Not quite. The phase one of the NSF SBIR grant really is a research proposal.

And our research proposal was to understand how well we can extract topics and discover topics from a whole set of RSS documents per se. RSS feeds are documents. So if we fed these RSS feeds into the system and ran a particular machine learning algorithm that we were proposing. It's called LDA, Latent Digital Allocation. How well would it perform? Would we actually get what we were proposing that we wanted to do? So that was our research proposal. And when you write an NSF SBIR grant, it really has to look like a research proposal.

That is a trick. I mean, it's not an engineering document. It's not, I'm going to build this piece of software. It's not a business plan. It's not a business plan. It's a research proposal. And I think your PhD helped you with this one. I think it might have helped with this one. Although I have no previous association directly with NSF. Basically, the advice I was given, and I looked at a few samples was you really have to write a research proposal for what you're going to do. But because it's SBIR, a small business, there is a small marketing section

in there. In fact, the first proposal application that you sent is only 15 pages long. Our original marketing section was seven or eight pages. And we couldn't fit the research proposal and the marketing plan into a 15-page document. And they're really strict about that. So we had to cut the marketing section down dramatically to about a page and a half or two pages and the first comments back from the review committee was your marketing plan seems weak and that was our chance to send the eight page or the 20 page document we had on our marketing plan and say look we actually have a marketing plan we just couldn't fit it in your 15 page application well that's always a great strategy is get them to ask you the questions you know how to answer yes but you have to make sure that you actually get asked the question and not get rejected right away yeah so you have to have

enough in there you said phase one and i know that the spi grants has multiple phases and the first one as you mentioned is usually a very researchy thing. And there are different amounts that they give you, buckets of money. And this is kind of a small bucket of money, sort of maybe one year's salary or a half a year's salary. It is 150K. That's the maximum you can ask with phase one. This is with NSF and every department, every agency is a little different. And phase two is what they call the commercialization phase. So when you're completed phase one and you have submitted your project report that shows a successful completion of the project and what you've found,

so it's like a research report essentially, then you have the opportunity to apply for phase two, which is considerably more money which is initially 750k with nsf enough to hire some engineers yeah that's correct so that's why we are you know hoping we're going to get the phase two we are waiting for the final word from nsf and that we will hire a couple of people so we have so far been approved by the program manager at NSF for our phase two proposal. We are waiting for final and last approval from the accounting and financial department. So that doesn't seem as bad. They seem like those are the people who are likely to just

make sure you're putting your receipts in the proper shoebox. I hope so. But they just take an awfully long time. So we've been waiting for that. So when did phase one start for you? We did phase one last year, January to June. And now we're June again. So it's been a whole year since. Right. So we had the opportunity to apply for phase two last July.

And if we had been approved, we would have gotten the money, say sometime in January of this year. But we deferred our phase two application to January of this year. So we won't hear back for the final approval until perhaps early July. And with your phase one, you said you got it in July, or I'm sorry, you got it in January. When did you apply for it? In July? June. In June.

June of 2012. It's awfully hard to build a company when money is six months away and uncertain. Well, I don't know if you've tried to raise money in the valley from venture capitalists. It can take a lot. It isn't lot fast either and that can take a lot longer than six months so six months is not bad from that perspective especially you know once you've gone past the phase one stage and you're applying for phase two it's considerably more money it's 750k that allows you to basically hire people take this product take this research that you've done to building a real product to being able to

commercialize that product. But it doesn't stop with phase two, does it? I mean, with NSF, it basically does. There is a little bit of matching post phase two, it's called phase two B, but that's not a lot of money. At that pace, too, is when they expect you to be able to either make revenue or raise more money, you know, from independent investors. So I'm embarking on raising more funds, basically, from angel investors and strategic money in the Valley in the July-August timeframe. So it's as though the government did your seed funding. Exactly. It does come with a cost. One of the reasons ShotSpotter didn't pursue SBIR grants is because they do retain some rights to your patents.

Not NSF. Not NSF? No, no. Oh, that's great. No, you completely own your IP, and they are there just to encourage research and to productize things, and this is part of the JOBS Act. Well, excuse me.

I think I need to go type some research. No, I'm sorry. Well, that's great. My argument with ShotSpotter was, if the government's going to take away your royalties for patents, they're not going to care if you have an SBIR grant. They're just going to do it. If it's military-related, they will just take what they need.

I cannot speak for the other agencies, so I'm not quite sure what the others do, but NSF does not take your IP. Well, the Department of Defense once didn't take your IP. It was just a scary paragraph in the document that said we might someday. But I think that's true when you become a U.S. citizen you have to promise that you will defend the country.

And part of me is like, why are you making people say that? And then the other part of me is like, why are you making people say that? And then the other part of me is like, yeah, I'm getting older and I'm out of shape and if they really need me to defend the country, it's kind of, wow, they're really down to their last barrel. So yeah, let's just go mark yes to that. I mean, if you consider that the country is being attacked more in the cyberspace than anywhere else, then they could use people like you.

That is true. Let's not give them any ideas. You know, but I think the issue with agencies like Department of Defense is they're worried that you'll develop such a key piece of technology that's important for the country's defense. They want to make sure they've covered themselves on being able to get hold of that technology from you. That was very much true. For gunshot location, that was very true. Right. And NSF is all about science research. And it's all about encouraging open research.

And it's an open community. So you mentioned you're going to get some additional funding and angels and VCs. And in Salt Lake and Valley, you could go to an entrepreneur event every single night of the week. Do you do those or do you focus more on the network you already have? I want to correct one thing that you said, though. I didn't say angels and VCs. I said angels and strategic investors. Okay.

So those are the keywords. So the difference is strategic investors are, for example, companies like Samsung, Nokia, Qualcomm. People who may want to buy you later. Not necessarily. They don't necessarily want to buy you, but they invest in you because they some of the companies that Qualcomm has been investing in, Qualcomm's not in the business of those companies, but they see value in strategically investing in those companies to moving the mobile market in those spaces. Ah, we want you to be a good company. We want you to succeed so that you will inadvertently sell more of our product.

Or enable others to sell more of our product. So it could be many levels of indirection here, right? So they have a strategic interest in you. And I see myself going to potentially the automotive companies, the tier one companies, even companies like Qualcomm and showing them what I have and showing them what I do and see if they're interested in that and in angels. So VCs, I don't think are super interested in what I do for two reasons. One, I'm a voice-based product,

but I'm using synthesized voice. And they've been bitten too many times in the valley in the past 20 years with companies like mine that have done synthesized voice-based products. So they're very wary of those kind of technology products. You can spend a lot of money trying to get your voice to sound good, and it will never sound good to everybody.

Exactly. The second reason VCs are typically not interested in what I do is because I'm trying to sell to the automotive market. And the automotive market, as you mentioned, is a slow-moving market. And that's a big company, you know. But once you get into a head unit, you're there forever. Exactly. So it's not a bad investment.

It's just a slow one. There is another aspect to the automotive companies today, right? They're all either integrating an iPhone or an Android. Most of them should be doing both, but yes. Most of them are doing both, but different car models, right? And I will be an app in these smartphones. And the car companies need to know who I am because they have a say in which apps on the iPhone and the Android

are going to be duplicated on this screen. How do you choose which car companies to talk to? Everyone. You have to go talk to everybody. Because you never know who will buy it, who will like it, and who won't. So I started to ask about the networking events. Do you go to a lot of those?

Are they important? Or are they just a way to spend an infinite amount of time talking to other people who are looking for money? You can spend a lot of time at these events. So I attend events judiciously. There are times in my startup career when I have been to three events a week. And there are times when I go to maybe one every other week, they are really useful for networking. And I have to say that people in the Valley are very, very helpful. They don't hold back. They give you advice.

They give you opinions. They give you introductions, actually, quite freely. That's the most useful. That's the most useful thing. And advice and opinions are useful, actually, quite freely. That's the most useful. That's the most useful thing. And advice and opinions are useful too. In fact, that's what I've learned over the last three or four years. So the biggest benefit from going to these events is you don't know who you're going to meet where and who will say, oh, you're trying to get a meeting with Qualcomm. I know a CEO of a startup who's been funded by Qualcomm. Let me introduce you to him.

And a startup CEO is usually too busy to have coffee with you, but not too busy to have a 10 to 15-minute phone call with you. So you get the introduction. You make a call to Mark or John or whoever and say, hey, John, I'm another startup. This is what I do. You have to have your two-minute spiel. Yeah, you got to get that elevator speech.

You got to get that elevator speech. And then the CEO says, Oh, sure. I'll be happy to introduce you. And he introduces you. And that gets you the first meeting where you get to do your 10-minute pitch to the investor. And this isn't bad for the investor either i mean they're out there looking for hot new companies too and having people find their way to them is actually you know it's kind of like a maze you have to get through the maze before you can start talking to the investors exactly and they'll only talk to you if you're introduced from someone they

know so you cannot send them a one-page executive summary or a business plan by email if you haven't actually met them or been introduced to them. They'll hardly ever call you back. So this is why I think these events are useful because you want to be constantly meeting new people, making contacts, and making introductions as well. So it's your job to also do this. So you met someone last week, you meet someone next week, this guy's doing something in healthcare, the guy that you meet the next week is looking for a company to be, you know, the marketing guy for a health company,

because that's where he came from. You introduce the two of them. They may never actually work together but they'll both remember karmic points for that exactly they both remember that you were a friend even though you're not really a friend and next time they'll introduce you to the guy who got yes it that's how it works that's how it works and i you know and i have to say i have been pleasantly surprised at these events in the valley how helpful people are. They, you know, are willing to talk to you, they're happy to introduce you, even if they've never met you

before. And if you are a startup, if you're thinking of a startup, if you're thinking of doing, you know, being an entrepreneur, I would encourage people to go to some of these events, you'll soon know which ones are more useful which ones are not so useful and which ones make the most sense for who you are because some of them are targeted exactly i mean you would almost certainly go to one that was targeted towards automotive things as opposed to the ones that are targeted towards health things exactly yeah exactly and you would you might go to some I might go to some that are targeting women entrepreneurs. I may go to some automotive stuff where it's very male. If you thought the embedded world was male dominated, you should see the automotive meetups.

And, you know, there are also meetups that are topic specific. So there's machine learning, for example, or there's Python, or you would go to... Moving on from the specific entrepreneurial meetups, meetup.com is kind of magical. I'm always surprised. I really could do... It's only my laziness holding me back. No, no, it's the long, hard day I had at work. That's what holds me back.

Right. But it is... And those events can be really interesting. I went to the sensor event in San Francisco before Solid. And I really went because Solid was happening that week and O'Reilly was sponsoring the sensor and I wanted to go start to meet some of these people. But, I mean, there were deep technical conversations. And then you'd move on in order, well, for me, in order to get closer to the beer. And then you'd have this interesting conversation

with an incubator founder. And just, yeah, I handed out a lot of, well, you called me, I'd like you to be on my podcast. So for the entrepreneur entrepreneur going to these would be part of your advice the technical ones and the entrepreneurial ones or try to focus well i think you could you can as you pointed out waste a lot of time going to these so you have to focus initially if you are a CEO of a company, then you probably want to attend some of the entrepreneurial startup focused ones. Go to ones where there are practice pitches, pitching sessions. So you get into the habit

of being able to pitch seamlessly without being nervous, you know, practice, practice, practice. And then don't go to any of these if you're not outward facing at that moment. If you're not fundraising, if you're not doing sales and marketing and biz dev, you need to really be focused on your product. Because this kind of stuff can be super distracting. And I found going to some of these as an engineer, the entrepreneurial ones as an engineer, people would talk to me and they would ask me about my product and I would say, yeah, I'm not really having one right now. I was so much less interesting than everybody else in the room, which totally feeds into my social anxiety as a whole. I'm a lot more popular at the technical ones where i can talk about technical things sure um to outward facing because it is an exchange as you were introducing them to

people they won't need to know they're okay they want that it's the introductions in the the networking part right i have to say though a technical person, the most in demand person at entrepreneurial events and meetups are technical co-founders. Yes, I was shocked. A ton of business and MBA types out there who are roaming the fields, if you will, looking for technical co-founders who can execute on their grand plan and their big vision. So it's a great place to sort of pick up people, you know, in the entrepreneurial sense. Yes, although sometimes they don't want to tell you what their idea is, or they're so sure that you should sign an NDA first.

And really, I think Michael Dell, ideas are commodity. Execution of them is not. Well, if there's anybody out there who's not telling you their idea, you should dismiss them right away. Because I've been talking about my idea everywhere all the time, and that's what gets me feedback and that's what gets me feedback that's what gets me introductions and the very first um by the way we have an automotive company today as a customer already i don't know if i mentioned this no you didn't so so you know can you tell us which one

no i cannot i'm sorry but they are having customers is fantastic. Having customers is fantastic. So they are using our API to build a demo product that's going to be demoed in the upcoming car. Are you going to get to go to the Detroit Auto Show and see your software? No, it's not going to be at Detroit. But it will be available sometime soon after that. And I think the first show is an internal demo. But I'm hoping that this is the first step,

and it is the first step to bigger and better things with them. Yes. It is all about the baby steps. Right. And they are a paying customer, even for a demo access to our API, which is really nice. Do you have beta users for your apps or not yet? No, we don't have any beta users today.

We did do a focus group study a year ago. Yeah, I think Jen Castillo was part of that. Jen was part of that focus group study. And I think what we really got out, we actually got a really good sense out of that focus group study on what people liked and didn't like with respect to topics and interests and sound and synthesized voice and usage. And what we're doing today is one of the direct, you know, results of that focus group study. So we came away understanding that even among a small focus

group study, the interests are all over the place. And so we have to cater to that wide interest group, we have to cater to the ability of people to have an easy way to get hold and make these playlists. Asking them to sift through 1000s of RSS feeds to create their playlist is not the best way to do it. No, no. My RSS feed for reading on screen is built up of many years, and then sometimes I go and I try to find more, and it's just too daunting. So we do it for you. So that's what this whole machine learning topic discovery is about.

We want you to tell us your interest area, and then we create it dynamically for you. Are you thinking about separating that from the voice and making it so that people like me could find more RSS feeds given either what I already have or with my stated interest set? So like I said, people can create their playlists either through our discovery process or the old way manually if they want. So you could do in our system, even on the app version that we have on the smartphone, you can do a search. So for example, if you really

like Huffington Post, you can say you can search for Huffington, it'll come up with all the Huffington Post RSS feeds, and then you can create your own playlist based on that. Is that what you were asking? No, no, I have a commute right now. I don't usually have a commute. So I don't have a commute so I don't have a lot of time that I listened to podcasts or time in the car really. So I was wondering if I could use it for the time I spend reading because I do spend a lot of time with my face in the screen reading news. So my question was really, is it separate? There's the voice generation and the reading to you, and then there's this machine learning intelligence part. And are they separate enough that if I didn't want the voice part,

could I get just the machine learning part and get a whole new list of feeds? Is that part of the plan, or is this really a contained unit and i should stop trying to take it apart i'm sorry we got so far off topic it's already taken apart so so for example think about how an automotive company might use our cloud it's a cloud-based service that we provide so most cars now are integrating text-to-speech into their head unit. So what they would get from us is just the discovery part and the feeds. And they would run it through their text-to-speech, for example, right? So they may have a favorite voice that they use all the time because they believe that's the familiarity of their users in the car like ford sync some people are bored to death with that but you know that's what the automotive company dictates is the way our drivers are

going to use our system this way so our system is already separate you know so that with the apis that the automotive developers will get they can either get the end product with the voice thing or they can just get the topics discovered or they can get the feeds so it's already there's already the ability to take it apart okay yeah i know that i can get some feed generation with like reverb but the last time i tried that it decided i was a girl and i only wanted fashion advice which that's annoying yes yes it is that it decided I was a girl and I only wanted fashion advice, which... That's annoying. Yes, yes it is.

Or shopping. How did you come up with this and why is it important to you? This is a lot of time and energy to sink into this product. So, I think I came upon this originally from the use in the car situation. Did you have a long commute? I didn't have a super long commute, but sometimes I would get stuck in traffic. And then, you know, I love NPR and I listen to it all the time. But there were two things that annoyed me a lot. It's one, a particular program, I might not be interested in that particular program.

Then I wanted to really hear about something that I just caught my eye at home and the dining room table on the New York Times front page. I had no way to actually get to that news. There's no way I could actually get to that when I'm stuck in traffic, which meant when I arrived at work, if it bugged me enough, I would spend a few minutes looking up that, you know, either go to the New York Times website or somewhere to just read up on that. Okay, they said something about Sangsam had this deal with this company. I really want to know more about that. There's no way I could get to it until I got back to my eyes-based reading method.

So that was one issue. The second issue is there were days when I didn't really want to listen to news at all. What I really wanted to do was relax, say on my drive back home from work. And I wanted to hear about gardening, which is one of my favorite things, or about travel. You know, we were going to South America, I want to hear about Ecuador, the Galapagos, I mean, something different. And there was really no way using the standard

radio in the car to get to information like that, or even Stitcher. You cannot get to that kind of information because Stitcher gives you podcasts, not necessarily current and relevant news or things that are happening in the world. And you can search, but what you end up with, I mean, we recently changed the show from making embedded systems to just embedded and and that was partially because it shows better in the titles but also things like stitcher now if you search for embedded my show comes up first and always first because it's the name instead of just a description. Right. And yet, if I was searching for, last week we talked about electric imps.

I guess next week we're talking about electric imps too. But if you search for electric imp, I don't know where you would find that. I mean, it's a product, but how would you find it in Stitcher? It would be very strange. But it sounds like with your product, I know there are a number of feeds that talk about Electric Gimp, and I know that they have blogs of their own, so I could add those if I wanted.

That's cool. If there are sufficient feeds, sufficient number of blogs on a particular topic, then those would get picked out into a discovered topic. And then when you searched for that, then you'd be able to find the associated channels. And then hopefully, you'll hear something about it. Yeah, for example, one of the big one of the topics that fell out quite recently from our machine learning runs, which we do every week, is robotics. And I was curious, I went and took a look at the robotics topic to see the channels. Oh my God, they were all robotics channels. You know, MIT and here and there and everywhere,

robotics channels. And so if you wanted to hear about robotics using our system, there it all was. So we have about... Boy, that would make my commute a lot more fun. Wouldn't it? Yeah. So that's the origin of this thing. I was frustrated and I said, okay, if I'm frustrated, everybody else sitting in this freeway with me are frustrated as well. What can we do to make this a better commute for people? And it's been a long, hard drive building the technology to get here but i think i'm i'm here now what was the thing that was unexpectedly hard

you know there are always these things you expect are easy and you expect some are hard were there things that you thought oh we'll just crank that it'll be easy but it turned out not i'm going to answer the question but not quite answer the question so i read a lot after we record i read a lot before i started my company i i was reading startup entrepreneurship venture blogs books everything nobody anywhere says it's easy I was reading startup, entrepreneurship, venture, blogs, books, everything. Nobody anywhere says it's easy, right? No one says it's easy. But before you actually start a company, it's an intellectual understanding of something that's difficult.

After you start your company, you really understand what it means for something to be hard. Because you have to really want to do this, to be able to get up every day and have the guy who worked the last two months for you suddenly stop answering your email and your phone calls, not show up, be able to go with the positives and the negatives, be able to continue to push through this. It's not for a few days, it's not for a few weeks, it's not for a few months, but it's for years. So that I think was, I wouldn't say it was a surprise because surprise is something that happens suddenly this unfolds to

you on a week and a month and a year basis a grindingly long basis sometimes but you can do this if you're still having fun and i'm still having fun well on that, I think that's a great note. Are there any last thoughts you'd like to leave us with? I would say to everybody out there, follow your passion. That's wonderful. My guest has been Radhika Teketh, CEO of Adjuvox. Thank you for coming over to talk to me. Thank you, Elle. Thank you for inviting me. Also, thank you to Christopher for producing the show. He wasn't feeling well, so he didn't join us on mic today. Unless he patched in comments later, in which case, could somebody tell me?

Hopefully, he will be on next week, though. And also, thank you for listening. Don't forget about that SparkFun gift card. I am quite looking forward to giving it away. Send your ideas for that or just say hello. Hit the contact link at embedded.fm or email us at show at embedded.fm or tweet to me at LogicalElegance.

I will have contact info for Radhika in the show notes, but I believe you can look at agivox.com, A-G-I-V-O-X.com. And the final thought for this week is Mark Twain, I think. 20 years from now, you will be more disappointed by the things you didn't do than by the ones you did do. So throw off the bow lines. Sail away from the safe harbor. Catch the trade winds

in your sails. Explore. Dream. Discover.