Utsav (5:00) Hey there listeners, TON122 is streaming live. (5:03) Today is going to be an insight-packed episode or stream, so stick with us till the very end. (5:10) Our first guest today is Achintya Gupta. (5:12) He's one of the co-founders of a company called RioDev. (5:17) They've announced a raise just last month, about $11.3 million raised in a round led by Elevation Capital. (5:24) It brings their total funding up to this point, about $15 million, and they're basically building a GTM stack for people who sell products to engineers. (5:37) And Achintya, welcome to the Offline Network. (5:40) Amazing, thanks. (5:41) Thanks, Dhruv. (5:42) Thanks, Siddharth. (5:43) Thanks for having me. Dhruv (5:44) Congrats on the raise. Utsav (5:45) Thank you. (5:46) So Achintya, I don't know how much justice I did to the description of the company. (5:50) Why don't we hear in your words and then we'll go from there. Achintya Gupta (Reo.Dev) (5:53) No, sure, Dhruv. (5:54) So I think it's great, I mean, the way you put it. (6:01) But we are like a verticalized AI GTM stack. (6:04) We are a GTM stack for companies that have technical buying journeys. (6:08) So like more than 40% of our software actually is purchased by builders, your engineers, your data scientists, your product managers, platforms, architects, etc. (6:19) And these are the people who notoriously are difficult to sell to because they don't want to talk to sales team, right? (6:27) But this is where the meat is, like this is where millions of dollars are made. (6:32) And companies like what Rio.Dev does is we look at millions of these signals out there in public and first party web and convert that into a sales intelligence and sales signals for our customers. Utsav (6:44) Fantastic. (6:45) So would one way to understand this, would it be fair to say that your product helps sellers understand how engineers pull the product and then use it once they've made the purchase? Achintya Gupta (Reo.Dev) (7:00) Yeah, so essentially these builders, developers might not be talking that much with the sales guys, right? (7:09) So the sales signals are not in sales calls, but it's out there like a lot of open source evaluations, right? (7:16) So many developers fiddle around with the product, with the trials of the products. (7:24) So the intent signals are in the code, in the either in the open source code we pulled out or in the product trials or somebody evaluating some other open source project, because if a developer has a problem, they'll be trying out a solution in open source. (7:40) So the signals of interest are out there in either public web or in open source data or even first party data for customers, like somebody coming on their developer documentation. (7:51) But no company looked at it from that perspective till now and said that, okay, can it be converted into sales signal? (7:58) And we were on the first ones which actually did that. (8:01) And yeah, that generates a lot of interest because it generates a lot of value. Dhruv (8:05) Did you realize this at your time with the API company Philo that intent tools are missing some of this? Achintya Gupta (Reo.Dev) (8:12) Yes, so I've been selling to technical buyers, same as first company. (8:16) It was like what we built there, what was a lending infrastructure product. (8:20) So we used to sell to CTOs, right? (8:22) Then Philo happened. (8:22) And that realization became stronger that, hey, you know, technical buying journey is fundamentally different. (8:29) Like, I mean, they have not been there very different than a B2B purchase journey. (8:34) And the intent signal is different and it has to be handled in a different manner. Dhruv (8:40) What are the top three things that the system looks at? (8:42) Like GitHub stars, CLI deployers, or docs being revisited? Achintya Gupta (Reo.Dev) (8:48) So actually, so it does look at open source, not just GitHub, right? (8:53) So GitHub is one way of looking at open source, but essentially you can get those signals in multiple other ways. (9:02) But yes, open source is one. (9:04) Then the second is the customer's first party data. (9:08) So their developer documentation or their product trials, right? (9:13) The third one, which is coming up a lot is agent evaluation. (9:18) So like I, as a developer, earlier was evaluating something and the customer wanted to know who is this. (9:24) And that's what Rio was giving. (9:25) But what's happened in the last couple of months is it is the agent of the developer that's evaluating. (9:30) So that intent signal is coming up a lot, right? (9:33) Like who are the organizations where developers have agents or where developers are building on their code on cloud or cursor. (9:41) And that's what we tell them. (9:43) And that's how their sales teams understand that this is where a silent evaluation is happening. (9:47) Because like our customers have a huge problem of an invisible sales funnel. Dhruv (9:53) Is that what you call pipeline fog as well? Achintya Gupta (Reo.Dev) (9:56) Yeah. Utsav (9:58) It appears that all of the signals you read eventually get distilled into what you guys call the developer knowledge craft. (10:05) So tell us more about that. Achintya Gupta (Reo.Dev) (10:07) Yes, exactly. (10:09) So like 90% of the GTM companies have been built on workflows. (10:19) Like you see many of them. (10:20) You can send automated emails. (10:22) You can send automated LinkedIn messages, etc. (10:25) Whereas the remaining 10% like us, they start with data, right? (10:29) So we wanted to be a data company first, a signal company first. (10:32) And we said that, okay, workflows can be built on top of it later. (10:35) So we realized that this entity or this persona, which is like a developer has not been sort of understood clearly. (10:47) Like for example, some of the biggest databases of contacts in the world don't give developer or technical buyers data. (10:56) And even if they give, then the way of getting the data of a technical buyer is very different than how a B2B software like a ZoomInfo will give. (11:06) Because if you are prospecting a technical buyer, you will look at their technical skills. (11:11) You will look at engineering teams. (11:13) So a typical engineering org has around 250 different teams, right? (11:18) Like if you look at just, you break it down into platform and mobile and CICD and so many others, right? (11:24) But not many had considered that. (11:26) So we wanted to create a graph of a developer's teams, their identity, their sort of, you know, their whole profile and map it to an interest graph that what this organization is interested right now. (11:41) So that is what we call a developer knowledge graph. (11:43) It's a living, thriving thing. (11:46) And we have, in fact, we have more than 100 million developers right now. (11:52) And yeah, that we just keep on growing it. Utsav (11:56) And the products that are being sold, are they being sold as subscription products, Achintya? (12:00) Or is it like, are people paying a one-time fee for them? Achintya Gupta (Reo.Dev) (12:04) Yeah. (12:05) Actually, we were very clear from day one, we will... (12:08) So it's a subscription product, but we do at least a year or more deal. (12:14) In the signaling space, actually selling monthly is very easy because somebody can just try it out. (12:21) You see huge amount of churn. (12:22) So for us, a sales happens when it happens for a year or longer, because that means we have been able to show enough value to the customer that they don't churn. (12:33) And that's something that actually I learned from one of the very smart guys. (12:37) He was in Clearbit, one of the OGs in this space, an intense space. (12:41) And that's how they used to follow that a deal is something like this. (12:45) Similarly, a customer becomes a customer once they have renewed. (12:49) So that's how you look at it. (12:51) But yeah, it is essentially a subscription product, no one-time thing, but at least a one year or longer deal. Dhruv (13:01) And what happens in the future when agents are actually the sellers and agents are also the buyers? (13:07) How do you assess for signal then and intent? Achintya Gupta (Reo.Dev) (13:10) Yeah, we are seeing that agents are evaluators. (13:17) So let's look at it from the buyer side right now, right? (13:21) So we are seeing that the agents are evaluators and the agents are, but the moment there is a serious purchase intent, the human or the person comes into the picture, right? (13:40) So in fact, in our product, there is a moment that is defined that, okay, from the agent, it is now moved to like a team evaluation or a person evaluation, right? (13:49) So that is one thing that we are seeing clearly. (13:53) The other thing we are seeing is when the agent is selling now, there the problem is, there are two kinds of agents selling. (14:01) One is which we saw for the last one year or so is it was not an agent sale. (14:11) It was like when the agent sale was happening on bad data, it was essentially irritating customers at scale actually, right? (14:21) Because essentially you have now a database, you have something and you can just mass spam people. (14:27) We wanted to build something that actually restricts it, right? (14:31) So hence, if you have the right data, the right context of the right evaluations, then actually the agent of our customers get trained to send the right messaging to the right people in the right time. (14:44) So for us, this whole picture that when the agent is evaluating and when the agent is selling, we feel that the right data and the right context actually bridges that gap that actually it becomes scalable. Dhruv (14:59) But say a company like NVIDIA and 11labs, like these are AI native companies. (15:04) So how are you embedding yourself in their workflows? Achintya Gupta (Reo.Dev) (15:09) Right. (15:10) So let's look at companies like NVIDIA or some of the bigger companies who are targeting these technical buyers. (15:17) There is... (15:21) So everybody has a GTM engineering team who we work with and everybody has sort of massive kind of, you know, scoring algorithms that these companies have. (15:35) And technical buying journeys play a very important role in it. (15:40) So it's like, you know, so for example, if you look at even companies like NVIDIA, which are essentially hardware companies, but software plays so much important role for them. (15:48) And for their sales teams to know that, OK, these are the organizations where the developers or the engineers are evaluating them is a very strong intent signal for them to go to the buyers and say that, hey, you know, this is who we are and this is what we are selling. (16:05) The second important thing in our industry is timing matters a lot. (16:09) Like if you enter very early and developers have just started sort of, you know, testing out or playing something, you don't want to get in at that point of time and it's too late, then they might have just built something internally. (16:21) So that's where sort of the right timing starts mattering for these guys. (16:25) That's where they use us. Utsav (16:26) Ritwik, just very quickly back to the agent question, is most of the work that agents are doing, is it pre-sales right now? Achintya Gupta (Reo.Dev) (16:36) On sell side? (16:37) No, actually. Utsav (16:38) No, on the buyer side, like, I don't know, reading the developer documentation, witnessing like, you know, basic demos. Achintya Gupta (Reo.Dev) (16:46) At the evaluation stage, yes. (16:49) So what we see is that if somebody is doing an initial build or an experiment, they might be sort of using these tools like cursor, but the moment sort of there is a strong evaluation, the developers will actually look at the docs themselves, will sort of probably be having a chat with the support engineers themselves or in their communities, they will be participating themselves. (17:14) So they are not relying 100% on the recommendations of the agent to actually build, right? (17:21) When there is something which is of production quality needed, the personal developer will like to have a say to understand what actually is happening, right? (17:31) And then once they are convinced, yes, then again, the agent will take over. (17:34) But that's where sort of a purchase has happened. (17:36) And then you are now moving it to a production. Utsav (17:39) Another question to understand would also be that if the product being sold as a developer tool, maybe the final buying decision is also a developer decision. (17:49) But if the product being sold is, let's say, enterprise software, very high contract value, then sure, the technical decision is the developer's decision, although, you know, CIO, C source, whatever's decision, but maybe the financial decision is the CFO's decision. (18:01) So do you have something to say with respect to that as well? Achintya Gupta (Reo.Dev) (18:05) Oh, yes, that's very true. (18:06) So that's where developer activities is a very strong intent signal, but our customers make sale on the buyer level. (18:14) So they will use RIO to understand that, okay, in this organization, let's say, Standard Chartered Bank, there are these teams which are evaluating the software and these are the people who are evaluators. (18:25) We give them the phase behind the evaluation and then we show them the hierarchy that, you know, this people or this team reports to this person. (18:31) And this is the right time now looking at the intent signals and the intensity of activities and evaluation that's happening, that you reach out to the buyer and have the discussion. (18:42) So it has to link to the buyer. (18:45) Developers are actually not the right person. (18:48) They are teams to be sold. (18:50) In fact, that's something we very strongly advocate. (18:53) Yeah. Dhruv (18:54) And in terms of quantifying all of this benefit to the companies and the logos that you're working with, do you have any numbers that you can share? Achintya Gupta (Reo.Dev) (19:03) So, actually, like, there is, we have seen massive pull. (19:09) Like, I mean, so just to share a few things, we grew 450% in the last 12 months. (19:18) Almost all of our business is global. (19:21) Like we decided to go global from day one. (19:26) And yeah, I mean, the other thing that we are seeing is that, like we started with SMBs and startups, and that's sort of what perhaps most of the companies do. (19:35) But now we are seeing increasing traction to our larger organizations who have massive sales teams and need the software. (19:43) So our thesis was that, I mean, the larger organizations who have, who are using some of these tools, like say Sixth Sense, et cetera, will be needing tools which are more new age. (19:54) So that has started happening in the last couple of months. Utsav (19:57) One question, like last year, when we saw, you know, for a lot of AI native companies, like revenue ramps that really look like hockey sticks, one of the things a lot of people are saying is, hey, this is all experimental ARR. (20:12) In other words, people are going to pull your product, use it for a while, only to replace it with another product. (20:19) Is that something you help companies measure and just give companies a reality check that, you know, while you'll see some early adoption and early revenue, it may not necessarily stick? Achintya Gupta (Reo.Dev) (20:34) We give them, so we give them the patterns of evaluations and activities that are happening. (20:43) So we don't help right now. (20:45) The use case we are not focusing on is what happens when the sales has happened, right? (20:50) Of course, you know, what you're saying, Dhruv, is right, the intent continues staying in the activities of developers. (20:58) So if they're not interested, if they're planning to churn, then they're looking at competing software, and that's another thing that we give. (21:03) But as a use case, we have not focused on post-sales till now. Dhruv (21:09) As a final closing question, Archanthia, maybe any good practices that founders can follow and this AI-heavy world of GTM that's upon us? Achintya Gupta (Reo.Dev) (21:19) Yeah, so it's a great question. (21:23) I think one of the best calls we took when there was so much pressure for us to build an AI SDR is we did not build an AI SDR and we built agents to actually give data to the AI SDRs. (21:37) And I think retrospectively, eight months later, we are just thanking ourselves for that decision because no AI SDR can actually work without good data and we solved that problem. (21:48) And the second thing we saw, so this is something that is now making sense. (21:53) The second controversial or tricky decision that we took was we went vertical. (21:59) And we said, we'll only focus on companies that have technical buying journey, not a horizontal market from day one. (22:05) What this has meant for us is that our signals are much more better. (22:08) The data, it's a vertical AI. (22:10) And once we look at it from that lens, it just keeps on getting better. (22:14) I think that's the next big bet that one year down the line, probably we'll be saying that, okay, we took the right call, which probably today looks slightly difficult to take. Dhruv (22:25) Wishing you very best. (22:26) And thank you so much for coming on our show. Achintya Gupta (Reo.Dev) (22:28) Sure. (22:28) Thank you. (22:29) It's a pleasure. Dhruv (22:31) All right, listeners, moving on to our next guest. (22:34) We've got Ajay from Coral joining us. (22:36) Ajay, welcome to the show. (22:38) Thank you. (22:38) Thank you for having me. (22:41) Can you introduce the company behind the name as well? Ajay Shrihari ( Coral Health ) (22:45) Yeah. (22:46) So we're Coral. (22:47) We have mission, we have two different sort of missions. (22:51) The first is to help reinvent process automation or to kill traditional RPA. (22:58) And what we focus on right now is helping process patients through the US healthcare system in less than seven to eight minutes of human time. (23:08) So typically it takes, in the US, it takes patients about 22 days on average to get contacted by specialty care providers. (23:18) Our goal is to help that happen in less than 10 minutes. (23:22) So if you go to your doctor's office or you get referred out to say an oncology center, typically that would take weeks on end to happen, that work. (23:32) We sort of make that happen. (23:34) And to pick up sort of page on a quick commerce and we say less than 10 minutes. (23:38) That's sort of what we do. Dhruv (23:40) But so how's the process getting shortened? (23:44) So I read that you built the company around sort of digitizing faxes, but also ensuring that when you cannot predict what it's written, the handwritten note that gets faxed around, you try to involve a human into the process as well. (23:57) And that brings your efficiency up to 99% for driving all of these exchanges between hospitals, departments, and clinics and stuff. Ajay Shrihari ( Coral Health ) (24:06) Yeah. (24:06) So the US healthcare system, unlike India, India is actually pretty well figured out in healthcare, actually funnily enough. (24:12) Whenever you go to your doctor's office, you go to say a Manipal in Bangalore. (24:17) I'm from Bangalore. (24:18) So it's very easy to actually just go in and just sort of get an appointment. (24:24) Here, there's a bunch of rules to get admitted. (24:27) There's a bunch of rules to follow to actually get the care you need. (24:31) And you need to figure out how much you actually have to pay out of pocket. (24:33) Healthcare here is incredibly expensive. (24:35) So what's interesting here is actually primary care, which is when you go to your doctor's office, your regular, say, pediatrician or the doctor you go to for the stiffles in some sense, that is like a huge loss-making exercise. (24:51) What that means is like a huge loss driver doesn't make any money. (24:54) People have been trying to reinvent primary care for years. (24:57) What makes money is actually specialty care, which is when you go to your oncology center, you get medical equipment at home, you go to an infusion center or so on and so forth. (25:09) That is sort of where the money is really made. (25:12) The problem is the transaction that happens between those two parties is never, I'm an engineer, so technical API call. (25:21) It happens via, even in 2026, a sort of fax machine. (25:27) And those spit out 125-page-long documents with 27 patients on it. (25:32) Multiple faxes might come in for each patient and multiple patients might come in for each fax. (25:38) You have, doctor's handwriting is notorious for being horrible, a lot of checkboxes. (25:45) So a checkbox could be four checkboxes, all scratched this way, a lot of chicken scratch signatures, things that traditional models are and are still sort of incredibly bad at reading. (25:58) And then you also have to qualify these on thousands of guidelines in each state. (26:01) You have to do benefits investigations, which should happen via phone calls. (26:06) It could happen via many, many different portals and forth. (26:18) That's sort of what it was. (26:20) So I think the problem here is you get 720-page-long documents. (26:25) The 17th page of Utsav might be in Dhruv's file. (26:29) How do you sort of reconcile all that? (26:30) I think that's sort of the absolute mess of a day that these people sort of deal with. (26:35) So one of our customers might get thousands and thousands of faxes each day in thousands of formats. (26:40) There's no one format. (26:42) And you'd have to hire hundreds of people to literally sift through that and do the work needed to process a patient. (26:49) That's sort of the idea. Utsav (26:53) Ajay, I mean, the U.S. healthcare system is notoriously complex, right? (26:56) And very often sitting afar, unless you've had a firsthand experience with it, you can't even understand the scale and the complexity of it. (27:04) But you're doing great work. (27:06) I'm sure you have peers who are doing great work. (27:08) Help us understand from a patient standpoint, how is the impact visible? (27:12) Also, what's the invisible impact before they arrive at the point of care, while they're at the point of care, and even after they leave the point of care? Ajay Shrihari ( Coral Health ) (27:20) I think we got inspired actually by one of my friends who's also from India, funnily enough. (27:25) He had just graduated from IT Delhi, came to study at CMU. (27:29) I was crashing on his brother's couch. (27:31) And the first day, it's like I had nothing to do. (27:35) This is around the time we were starting. (27:37) So when you're starting, you don't have much to do. (27:41) Went to a primary care doctor. (27:43) The person spent four minutes with us, never looked my friend in the eye and said, you got to go see a specialist. (27:49) I said, okay, where's the sort of specialist? (27:52) They said, oh, I'll refer you to my gastroenterologist who was on the Upper West Side in New York. (27:58) I said, okay, when will I find out? (28:01) I said, whenever she reaches out to you. (28:03) She reached out to my friend in 14 business days. (28:06) So that's about 18 real days. (28:08) I said, hey, you can come in tomorrow. (28:10) Everything's sort of ready. (28:11) We go in there. (28:14) She spent seven minutes with us and said, you have to go get a scan. (28:17) I said, well, shit, I could have probably known that with ChatGPT, right? (28:21) I should probably go get a scan. (28:22) That's pretty obvious. (28:24) And that took another six days to happen. (28:26) In the meantime, my friend went all the way back to India, went to a doctor's office, got a specialist to see him, got a scan done, realized his gallbladder was about to burst, got operated on and showed back up to the U.S. eight days later. (28:40) And I still hadn't figured out how to get the scan, right? (28:43) So that's sort of the story. (28:44) If you ever go to like in the U.S., you go to any dinner table conversation and ask, hey, how long did it take to go to the next appointment? (28:52) You say, oh, I have to go to this appointment. (28:54) It's hugely problematic. (28:55) I have to figure out how to get there. (28:57) The burden is sort of on the patient in some sense to sort of orchestrate their own care, right? (29:05) Which is actually fairly different from other countries that I've sort of been in. (29:11) So in this case, it's a magical experience. (29:13) If you can, you go to your doctor's office, you're getting referred out, which happens to most, I would say 65 to 70% of the patients in the U.S. healthcare system will typically get referred out. (29:25) Instead of that taking say a 22, 23 day experience, you know how much you have to pay, you know who's going to take care of the next part of your care journey. (29:36) You know where to go in some sense. (29:39) And all of that, as soon as you walk out of one of your doctor's offices, I think that's a great experience. (29:44) It's a consumer-esque in some sense experience in an industry that's traditionally been very bureaucratic. (29:53) I think there's a lot of anxiety in all that, in all honesty. (29:56) Because of the fact that no one knows. Dhruv (29:59) When you looked at this problem statement or discovered this problem, like were you going to the customer or the hospital and the vendor or the clinic and saying that I'm going to digitize all of this at once? (30:10) Or did you go by vertical or vertical? (30:12) Because I read, I think you went after medical equipment first. Ajay Shrihari ( Coral Health ) (30:16) Yeah, I think the U.S. healthcare market is not one market. (30:19) It's a trillion dollar market, right? (30:22) In some sense, probably the largest market in the U.S. But out of maybe the two trillion spent on healthcare, you can actually divide that into not one market, maybe hundreds of sub-markets in some sense, each that operate fairly differently. (30:41) So it feels like sometimes you're building a different company, right? (30:46) So when that happens, when you're a young company, you want to solve one problem at a time. (30:52) You don't want to take too much on. (30:53) And you want to do it in one vertical. (30:56) So we actually went after durable medical equipment because these folks, I don't know if you've seen Shopify merchants in some sense, like people who build on Shopify around the globe. (31:06) We actually felt like these people sounded more like Shopify merchants, more business-minded than traditional healthcare providers would. (31:14) And they're the last sort of part of the care journey. (31:16) And their care continuum, they're like the last node in the graph. (31:21) So these people would get thousands of these documents, and 100% of this would be via faxes. (31:29) And they'd make these decisions a lot faster. (31:32) Over time, you say, okay, now that I've processed, say, half a million patients, a million patients, now 1.8 million patients per month through the platform, you've earned the right in some sense to go after every other vertical because you're now the vendor of choice who's now doing this for so many patients in the US healthcare system. (31:48) People will typically buy you a lot faster. (31:51) In other verticals, like even health systems, for example, where traditionally the buying cycle would take five to six years, here it would be less than six months. (31:59) So that's kind of why we chose to do one at a time. Utsav (32:05) Anujay, everyone keeps citing this one big number, which is $5 trillion in annual spending, I think, in the US healthcare market, which is saying something, right? (32:14) It's the GDP target we've set for ourselves as India. (32:17) But where's the revenue opportunity in all of that for companies like Korn? Ajay Shrihari ( Coral Health ) (32:25) I think right now, the biggest shift is, I think the biggest cost, actually, in the US healthcare system is the labor cost. (32:35) The revenues, in some sense, it could be administrative and clinical, right? (32:40) You have, I'll give you a great example. (32:44) The biggest cost center for most of our customers is administrative in some sense. (32:49) Like for a $2 billion, $3 billion revenue practice here, just for, say, reading information off of documents and putting it into systems, I've seen people hire close to 500 to 1,000 people. (33:07) Now, these people could be in the US, they could be in India, they could be in the Philippines, whatever it is, whichever country. (33:13) If you look at that number on the balance sheet, it grows exponentially each time you grow. (33:22) So these are growing tremendously as a business. Utsav (33:24) These are people who show up to work in scrubs, but just do data entry all day. Ajay Shrihari ( Coral Health ) (33:29) Pretty much at this point. (33:33) That's where all the time goes. (33:35) The second thing is, most, you asked where do startups fit in, most ERP and EMR systems haven't grown. (33:43) So the story of the ERP actually here is, there was a high-tech act in 2007, which said, let's make everything paperless. (33:50) They went from the filing cabinet to the cloud directly. (33:54) And in some sense, it was subsidized software, which is, someone is forcing you to adopt a technology. (34:01) And that creates this sort of boom of EMR and ERP systems and Epic and other companies grew very, very quickly in the 2000 and 2010s. (34:11) But these things were not a great improvement over the incumbent. (34:15) This is not great technology. (34:16) This is not great software. (34:18) I don't know if you've ever opened some of these EMR companies or seen what this looks like. (34:21) It looks like a Windows 80 application. (34:24) There's a great, and then once, how do you price? (34:28) You charge on per seat, right? (34:30) So that means every time a new person gets added in the healthcare provider's office, you make another $300, $400, $1,000 per month. (34:38) It's an incredibly expensive tool. (34:41) There's a great opportunity here to sort of reinvent what that looks like and build that 10X improvement now, because you have AI and you have a lot of these other things to build great systems of action. (34:52) So you're actually gonna see, and you're actually seeing across, say, healthcare, specifically, you're seeing the boom in voice. (34:57) You'll see it in PriorAuth. (34:59) You'll see it in scribing. (35:01) You'll see it in intake like us. (35:02) And each one of these verticals, you'll see a new player that essentially is trying to become the system of action and do a lot of the work. (35:11) And most of it is sort of built on the fact that the labor cost, it doesn't have to be what it is. (35:16) And you can give the time back to a lot of these providers. Dhruv (35:20) Ajay, as a final closing question, congrats, firstly, on the 12 million that you've announced earlier this year. (35:25) So how is growth looking like for you? (35:27) And, I mean, congrats on getting two of the biggest names in the Indian venture world, Lightspeed and Z47. Ajay Shrihari ( Coral Health ) (35:34) Yeah, growth's been good. (35:35) I think, actually, just this month, we 3X revenue. (35:38) So that's pretty good. (35:39) And I think we're gonna double it again. (35:42) I think what we're seeing, actually, is traditionally, verticals like healthcare have been slow to adopt. (35:50) It's probably the slowest part of technology. (35:53) I think it's probably adopting as fast as legal right now. (35:57) Once you build a great experience, I think people will jump on the bandwagon a lot faster, even in verticals like healthcare. (36:05) We're seeing half a million dollar deals close in three weeks. (36:09) It's that quick. (36:11) Because the problem is so acute. Dhruv (36:14) I mean, the problem is so acute, but why has nobody figured this out? (36:17) I mean, you just plug into a fax machine and digitize that, right? Ajay Shrihari ( Coral Health ) (36:20) It's not really that. (36:22) It's not just that, right? (36:23) Then we expanded into, how do you do benefits investigation? (36:26) Then we did qualifying a patient. (36:27) We built patient communication. (36:29) We built... (36:29) That was the entry point. (36:30) We actually are a full-stack provider now. (36:32) We go all the way to submitting authorizations and more, which is probably the biggest cost center within the US healthcare system. (36:39) So I'll give you a good example. (36:41) One of the biggest RPA companies in the world is one point close to 2 billion in revenue. (36:48) Maybe about 894, 900 million might come from just referral management and prior odds, which is kind of what this vertical is. (36:58) This problem is very, very acute, and it goes... (37:01) The person, you can sort of start with one incredible wedge, and then you can build everything. (37:05) You own the right to do every single workflow downstream. (37:08) I think that's sort of what we're seeing. (37:10) And hence, people are buying everything in some sense, and they're committing long-term, even to a young company. Dhruv (37:18) And how many customers are you working with now? Ajay Shrihari ( Coral Health ) (37:22) I don't know the exact number of customers right now, but I think close to 50, 60, I think, is where we'll end up by the end of the month. (37:30) But these are all very large enterprise sort of customers that all pay us. (37:35) Our minimum contract value is, I think, 129 or something like that, thousand dollars. (37:40) And our largest... Dhruv (37:44) All right. (37:45) Wishing you and the team Coral the very best. (37:47) Thank you so much for coming on our show, Ajay. (37:49) Thank you, Praveen. (37:49) Have a good one. Utsav (37:51) All the best, Ajay. Dhruv (37:52) All right, listeners. (37:53) We're moving on to our final guest today. (37:55) We've got Shilpa from Lemniska. (37:57) Shilpa, welcome to the show. Shilpa Nargund (Lemnisca) (37:59) Thanks, Atsav. (38:00) Nice to be here. Dhruv (38:02) Is that your office floor? Shilpa Nargund (Lemnisca) (38:04) Yes, we are incubated in Scalar School of Technology. Dhruv (38:08) Okay. (38:09) So why don't we introduce the company? (38:11) And I would love to know what the name also means. Shilpa Nargund (Lemnisca) (38:15) Yes. (38:16) So I'll start with the name then first. (38:18) So Lemniska comes from the word called Lemniscate, which is essentially the infinity symbol. (38:24) So yeah, we are infinite possibilities if we read between the lines. (38:30) And what we're building is for the biomanufacturing industry. (38:34) Right, both Pushkar, who's my co-founder, and I come with a lot of experience in the biomanufacturing space. (38:42) And we saw very often that a lot of products fail before they get to manufacturing. (38:47) And that is the space we are working on. (38:50) So if I had to describe this in one line, it's like if there is any product being made in a bioreactor, we'll improve its manufacturability. (39:00) And I can spend maybe a minute or so elaborating on that. (39:05) So what is a bioreactor? (39:08) It's very similar to a fermenter that makes pure or ethanol. (39:12) And now it can make a lot more things. (39:16) I mean, if you've been following the gold rush into the AI drug discovery, most of these drugs will have to be produced in a bioreactor before they reach a patient. (39:26) But it's not just drugs that the fermenters make. (39:29) They make food ingredients like vanilla essence or saffron essence. (39:34) All of this is made by genetically engineered microbes or cells within these tags. (39:41) And so basically, you can make paints, dyes, food ingredients, cosmetics, nutraceuticals, you name it. (39:50) And all of this is called the bioeconomy. (39:53) And basically, India has been putting or the Indian government has been giving a lot of impetus for this sector, especially because you might have heard of the bio E3 policy, right? (40:06) So they really want India to become a global biomanufacturing hub. (40:12) And that ties in very strongly with our pitch for Atma Nirbharta. (40:18) Because biomanufacturing is the only technology where you can take local raw material and convert it into a product that you want. (40:26) And very often these local raw materials could even be waste. (40:30) So agri-waste and so on. (40:32) So that's the whole piece about what bioreactors, what products, what's the bioeconomic space. (40:39) I can pause here or I can go into why Lemniska chose this space and what's the problems with manufacturability and so on. Dhruv (40:47) I think your background will be very interesting to know. (40:49) How did you land up into this? Shilpa Nargund (Lemnisca) (40:51) Right, so both Pushkar and I, we are chemical engineers. (40:55) So this is interesting. (40:57) After we finished our undergrad and we both got our PhDs in the US, that is the time when a lot of chemical engineering departments were changing their name from chemical engineering to something like a biomolecular engineering or biochemical engineering. (41:13) So suddenly designing a cell or a biological system to make something was becoming a big rage. (41:21) This is about 15 years ago. (41:23) And that's where we landed. (41:25) We got to work in some very exciting labs. (41:28) And since then, we've been in the bioprocessing industry. (41:32) And yeah, I mean, how did we end up making Lemniska? (41:35) Quite frankly, we are chemical engineers and this is all we know how to do well. (41:39) So that's why we're here. Utsav (41:46) All right, Shilpa, I'm going to ask you a question without knowing the right terms to use here because both Siv and I are total aliens to your industry. (41:53) But whatever goes in as input into the bioreactors, firstly, is there a term for it? Shilpa Nargund (Lemnisca) (42:00) Yeah, bioreactor or fermentor. Utsav (42:03) I'm saying what goes in? Shilpa Nargund (Lemnisca) (42:05) Oh, what goes in? (42:06) Oh, okay. (42:06) What goes in is a feedstock. (42:09) Right, it's a feedstock. (42:12) I mean, it depends which industry you're talking about. (42:14) If you're talking about biopharma, we can just call it nutrients. Utsav (42:19) Yeah, the feedstock or the nutrients that you're working with at this time, also because you're a young company, are you working with tried and tested candidates that are ready for production scale or are you still working on what in the physical world we call prototypes? Shilpa Nargund (Lemnisca) (42:38) Right, so we are a seven-month-old company, right? (42:41) And I'll tell you what we are building and then I can come to what we are working on right now because that will make more sense. (42:47) So I stopped at what the space, but what's really the problem with manufacturing, right? (42:52) And I can just go back to take the example of the AI-designed drugs. (42:57) Now that's just step one. (42:59) You decide what molecule to make, right? (43:01) But then there's a lot more you need to do. (43:03) You need to actually improve the process to make a high enough yield of this product so you can actually at least serve a few patients. (43:12) I mean, this is regardless of the clinical trials. (43:15) Even if the clinical trials go through, this is what you will have to do, design this process and scale it up. (43:21) So when drug companies are working now, they're working with say one milliliter of volume. (43:27) So this is all liquids with cells and their nutrients floating around. (43:32) So we are always talking about volumes in bioreactors, right? (43:35) So they're working in one ml. (43:37) If they want to make it commercially, they have to scale it up to 50,000 liters. (43:42) If it's not a pharma drug, then you have to scale it up to one lakh liters or more, right? (43:49) So there are a lot of engineering problems as you move from discovery. (43:52) So it becomes less biology and more engineering as you scale it up towards- So you're saying it's not linear basically, like you can't scale it one ml. Dhruv (44:02) I mean, you can't multiply everything by 50x and expect the output. (44:06) So how many projects that you run in a lab actually make it, I mean, suppose you want AI to come to a AI design drug and how, I mean, it works in the lab, like when you do one in 10, 15 pills, but if you want to make it for commercial grade, like what really makes and how much of it makes from lab to actual market? Shilpa Nargund (Lemnisca) (44:25) So that's a staggering number, actually. (44:28) I mean, not just considering drugs, but all bioproducts in general, if you consider what has been tried in the lab, only 10% of that has translated to manufacturing. Dhruv (44:39) Okay, let me ask you another interesting question. (44:41) Just, I mean, leading this on is, do you know any interesting projects that have worked in the lab, but not made it outside? (44:48) Like, I mean, solving cancer or, I don't know, solving something that works in a lab, but we've not figured out how to make it commercially yet. Shilpa Nargund (Lemnisca) (44:56) Well, a lot of things. (44:58) Okay, I mean, when we say lab, let's be clear. (45:01) If you're talking about a few milliliters, there are tons of, so saffron, which is grown in flowers, that has been made in the lab. (45:12) It's not manufactured yet. (45:15) There are so many biopolymers, bioplastics that show a lot of promise in the lab. (45:21) But haven't been manufactured. (45:22) There are drugs even that, I mean, with drugs, it's a bit of a different problem because there your profit margins are so high that even if you can get it manufactured at some low yield, you're good to go. (45:36) But I do know a case where there was this company called Zenzime in the US, where they produced something at 150 liters. (45:44) And they were providing this to infants as patients. (45:48) And they wanted to scale the same thing up to 2,000 liters and provide it for the same indication to adults. (45:55) But the moment they moved from 150 liters to 2,000 liters, the product profile completely changed and the FDA asked them to go through clinical trials again. (46:04) So that's how acute the manufacturability problem can get as you move up the scale. Utsav (46:14) Shilpa, we'd also love for you to talk to us about, you know, just sheer data complexity in the work that you do and the intelligence platform that you've built. (46:22) But before that, who makes fermenters in India? (46:24) Do we have indigenous manufacturers of fermenters as well? (46:27) And what does the largest fermenter in India look like? (46:31) Would you know? (46:32) Have you seen it? Shilpa Nargund (Lemnisca) (46:34) Well, there are a lot of people who make fermenters in India. (46:37) I mean, well, fermenters also come in a lot of different classes. (46:41) If you're making ethanol, it's a fairly cheaper fermenter versus if you're making monoclonal antibodies, which are these drugs, right? (46:48) But there are people who are making fermenters in India. (46:50) But obviously, we also import if required. (46:54) The largest I have seen is, I think, a 2,000 liter. (46:58) But I have heard of a 500,000 liter, which is five lakh liters. (47:03) We're talking to a customer who is going to build a six lakh liter fermenter, one unit. (47:09) So that would be, I don't know, maybe like a 10-story building. Utsav (47:12) Like custom commissioned projects, someone comes and makes them for you. Shilpa Nargund (Lemnisca) (47:15) Yes, yes, yes. (47:17) So maybe going back to the piece that you asked, right? (47:21) So what are we using? (47:22) What kind of feedstock we are using? (47:24) We are using prototype feedstock right now. (47:27) So we have our own wet lab. (47:29) So maybe that segs into what Lemniska is building, right? (47:32) So we are building both models. (47:34) So that's the digital stack and our wet lab to validate these models. (47:39) So right now, we have a wet lab that can go up to 100 liters. (47:43) We have two programs running there. (47:45) And I think the whole philosophy of Lemniska is that we can model these things. (47:53) And models are not so commonly used in the biotech industry as you would see in the oil and gas or other process industries. (48:01) That's the white space we're trying to fill. Dhruv (48:05) And you're doing software on top of this. (48:07) So you're doing your own wet lab. (48:08) You're doing software on top of customer data to come to a solution basically for your customer. Shilpa Nargund (Lemnisca) (48:14) Absolutely, yeah. Dhruv (48:16) You're seven months old, but where do you see the first set of applications for your software coming out? Shilpa Nargund (Lemnisca) (48:23) Yeah, so what we did was we built our own lab and we figured out what we need to make it more automated, more agentic, if you will. (48:35) I mean, we're trying to harness everything as AI as far as possible to make things efficient. (48:40) And what we came up with was a framework of models, obviously. (48:45) And when I say models, these are not AI models. (48:47) These are physics-based models. (48:49) So we are essentially solving a bunch of differential equations in these models. (48:54) That's what explains how these microbes will grow or produce or consume a given feedstock, right? (49:01) And so we built a software solution which we are already talking to customers about. (49:07) It's called Trellis. (49:08) And in that, it's what helps a bench scientist go from doing experiments in a very trial and error ad hoc manner to using models to guide these experiments in a structured manner. (49:23) So that's our first thing that goes to customers. (49:25) But there's a lot of things coming ahead. (49:28) Obviously, we're only seven months old. (49:30) So the way we would like to see this business grow is both in the software stack, but also in the wet lab stack where we are building a data flywheel. (49:40) We are talking to select customers to give us their projects that we develop up to 100 liters and then help them scale that up further with partners. Dhruv (49:48) It's like a flight simulator for bioprocessing. Shilpa Nargund (Lemnisca) (49:50) Absolutely, yes, yes. (49:52) That's the word we use as well very often. Utsav (49:55) I should probably be zooming out a bit towards... (49:57) I mean, when we just got started, you told us how much of a priority this is even for the government, right? (50:00) To just catalyze a bioeconomy. (50:03) So at this point in time, where do we stand? (50:07) Where do things stand? (50:08) Which countries actually have the largest bioeconomies right now that we are maybe collaborating with? (50:13) And what is the path to becoming one of the largest ones, if not the largest ones? (50:18) Also, which are some like just some no-brainer type feedstock ideas that everyone knows about. (50:24) And maybe you could educate us and our listeners about as well. Shilpa Nargund (Lemnisca) (50:27) Sure, sure. (50:28) So I'll start with the bit about where India stands in the biomanufacturing field, right? (50:34) We actually do quite well. (50:36) We have... (50:38) You might have heard of biologics and biosimilars. (50:41) So if you're talking about chemical drugs, they are generics. (50:45) But if you're talking about biological drugs that have gone off patent, those are called biosimilars. (50:50) And we do a very good job with that. (50:52) Biocon, Dr. Reddy's, Enzine, Lupin, Cipla. (50:55) There are a lot of companies doing that, right? (50:57) So we have the biomanufacturing base. (51:00) We have the talent pool. (51:02) Even in non-pharma, we have a lot of good companies doing that work. (51:06) So I see that this policy is... (51:12) And we've been talking to a lot of startups as well who are entering this space. (51:16) So there's a lot of scope within India, but the market for the products that come out of this, it is still outside. (51:25) And I think the feedstock, I think you're very interested in the feedstock. (51:28) So the common feedstocks, right? (51:29) Quite frankly, it's something like the gases or molasses that come out of sugarcane. (51:38) And that's pretty much how Pushkar and I started. (51:41) We thought, okay, we come from the sugar belt, both him and me, between Maharashtra and Karnataka. (51:47) And we're like, we really need to upgrade what's happening with sugarcane now. (51:51) The best you can do with sugarcane is sugar or ethanol right now. (51:54) With the same process that makes ethanol, there's so many other things you can do. (51:58) So that was one of our starting points as well. Dhruv (52:03) Any final closing question, Dhruv? Utsav (52:06) No, I'm mostly through. Dhruv (52:09) I think it's been quite educative speaking with- I mean, maybe like a forward looking one, as a closing one, if AI is able to scale at least simulations in terms of software, like simulating the real world and stuff. (52:22) Right now we're restricted by somewhat physics, I believe, right? (52:25) Is that safe to assume? (52:29) Presume? Shilpa Nargund (Lemnisca) (52:31) Yes, yes, it's understanding the physics. Dhruv (52:35) And, okay, maybe I'll say into how we see- I think the thing that I was leaning into was asking, will AI become sufficiently advanced and have so much real world data that it will be able to predict at scale some of these things? Shilpa Nargund (Lemnisca) (52:51) So I think the LLNs that we work with on a daily basis, like Claude and Chachipiti, may not. (52:58) Already with the product that we have, they do not do as well as what we have built because our products are constrained by real tools that are based on physics, right? (53:09) And so I do see what we are also tending towards what are called as world models, right? (53:15) I mean, if we can codify a lot of the physics and the chemistry in there, then building these world models with data that we can generate synthetically or get through our own lab, that is the way to go. (53:31) That's where we are looking forward as well. (53:34) But AI alone will not do it. (53:37) That's my core belief. (53:38) You're going to have to put in laws of conservation of mass, laws of conservation of energy and all of the good things we know from like centuries. Dhruv (53:48) All right. (53:49) Wishing you and the team a very best, Shilpa. (53:51) Thank you so much for coming on my show. Shilpa Nargund (Lemnisca) (53:54) Thanks for having me. (53:55) It's been a pleasure. (53:56) Thank you. Dhruv (53:58) All right, Dhruv, you and me. (54:00) Before we break for the weekend, I think there's a lot of personnel changes that have happened, right? (54:04) Google put out something as well. (54:06) Update us. Utsav (54:08) Yeah, what just happened? (54:09) So I think we'll start with Google. (54:14) So Jeff Dean, who was, I think, for developers around the world, like the biggest icon ever. (54:20) Him and his longtime collaborator, Sanjay Gemawat, they were both level 11 engineers at Google, which is like the highest level. (54:28) They've left Google after, I think, like a nearly three decade long career. (54:33) They've announced a new company, which is set up as rather a new organization that's set up as a public benefit corporation called Discovery Loops. (54:41) And they're leaving. (54:43) And Jeff Dean used to be chief scientist of Alphabet. (54:47) So that's the number 15, by the way, at Google. (54:52) Yeah. (54:53) It's crazy. (54:54) Like that duo has worked in every product that we've used like ever. Dhruv (54:58) So as you can see, I think that slide, which is like characteristic of Google, I think everything Google sort of looks like that. Utsav (55:08) Yeah. (55:08) So that's one big piece of news. Dhruv (55:12) The other big- What is the company doing? (55:13) Like, I'm going to do a small readout. (55:15) I think so. (55:15) Yeah. (55:16) I mean, so it's a public corporation whose entire premise is that the scientific method is an algorithm humans have been running by hand. (55:24) Hypothesis, experiment, result, repeat. (55:26) But he's now betting with this new company that you can run thousands of these loops in parallel, starting with machine learning research itself. (55:32) So improve the machine and the code and then move into chip design, drug discovery, and clean energy. (55:37) And Alphabet is a founding investor and a cloud partner. (55:40) So they will partner with Jeff and Sanjay as well. (55:44) So exciting. Utsav (55:45) Basically, the machine that builds the machine, which is what anyone like Jeff Dean, that's the kind of thing that they're working on. (55:52) So it's quite something. (55:55) And I think the round was either led by or saw participation by Khosla Ventures as well. (56:00) And Vinod Khosla put out a tweet saying that you guys could have had anyone you wanted. (56:05) Imagine Vinod Khosla saying that you could have had anyone you wanted on the planet. (56:08) So thank you for choosing us. Dhruv (56:11) Vinod Khosla has had quite a run, actually, by the way, in this world of AI investing. (56:15) I think it was very early, of course, with OpenAI. (56:17) But of course, many others that he's done. (56:20) Lightspeed too. (56:21) Lightspeed, Kleiner-Burkins, then Alphabet. (56:23) Like, I mean, amount and valuation are undisclosed. (56:25) But it's still open, even at announcement. (56:27) So let's see. (56:29) Yes. Utsav (56:29) And then Google's Nobel laureate, you know, in-house person, Demis Hassabis, is also, well, some say he stepped down as CEO. (56:40) Others say he stepped up as chairman. (56:41) But he's basically going to take a step back from running the day-to-day at DeepMind, which I believe he wasn't doing for the last one year anyway. (56:48) So there are some changes there as well. Dhruv (56:51) And DeepMind's CTO, I think, Corey, who was with them for 13 years as CTO, is now taking over to lead this day-to-day. (56:58) Alphabet stock fell 4% because of this. (57:01) 4%? (57:03) Elon Musk's company changes as well. (57:06) Nikita. (57:08) Nikita was, I mean, with Elon's company X, which is now, of course, merged with SpaceX. (57:14) He was the head of product there for just over a year. (57:17) And he's moving out. (57:18) He will only be an advisor. (57:20) And SpaceX did their first public earnings and gave us a first-year look at the ads business, which is 367 million for the quarter against 1.08 million in last comparable quarter for Twitter in 2022. (57:33) Because we don't have the stats in the middle. (57:35) So ad business has definitely taken a hit. (57:37) But I'm guessing that they make up for it in the subscription business that's come up. Utsav (57:41) Yeah. (57:42) Rumor has it Nikita, he's leaving to focus on his sleep because what he was doing is basically full-time, 24-7, non-call kind of job. Dhruv (57:51) Dude, like people shouting at you. (57:52) You can never keep people happy. (57:54) And since he was fairly public, he was involved. (57:57) And I think just, I mean, this public town square that X has become, you just like literally cannot keep everyone happy. (58:03) And there is this screenshot that was going on that he has this consulting call, which is apparently $15,000 an hour to, I think 30 minutes or $15,000 an hour to just chat with him if he even accepts. (58:16) So imagine. (58:17) All right. (58:18) I think that's it from us. (58:19) We'll see you on Monday. (58:20) Have a wonderful and safe weekend. (58:21) Thank you, everyone. (58:22) Bye-bye.