Audio & Transcript

Welcome to the Superconnector podcast. I am Matt Joseph, your host. We've got another round of startups to go through today. Some really exciting ones, especially in artificial intelligence, robotics and other industries. Let's jump into it.

First up. Deep Night Next generation night vision. Building night vision with AI. That sounds interesting.

Seeing in near total darkness is usually solved with specialized and expensive hardware. Like the $3,000 image intensifier tube. Solutions like thermal and infrared cameras are extremely expensive and power hungry. Digital cameras are still unable to perform in near total darkness like moonless starlight.

Very artistic of them. We leveraged state-of-the-art digital CMOS sensors, high performance, mobile SOCs, and edge AI processing to build real time on edge night vision that enables devices to outperform the state-of-the-art in night vision technology.

Whew. That is a mouthful. I think using less jargon would help that framing. But I see what they're trying to say here. Basically, they're using a bunch of cool technology to create night vision. And they're letting you know that they're insiders and they're smart by using all this jargon. But you can get there with more approachable language.

Okay. So we've got a video here See what we got.

they're showing just how well it works. They have basically two cameras that are using this technology and you're seeing what it looks like in real darkness versus what it looks like otherwise. That's really cool.

The team. Thomas and Lucas grew up in New Jersey together and stayed in touch. Thomas went to UC Berkeley and studied EECS before hopping from Lawrence, Berkeley national laboratory to Lyft, to Google working primarily on machine learning systems. Lucas went to Cal poly SLO and specialized in computational photography, researching edge computer vision problems at Meta and the Google pixel team. Well, so these are some very accomplished founders, very smart childhood friends. The first night photo I've seen at Y Combinator, very on brand night vision team. I like this great team. we are supported by the us Army's night vision and electronic sensors directorate and have received an LOI from Anduril. Industries.

Well, that's incredible Anduril basically the hottest company in defense right now. And the army is obviously a giant customer if you can get them.

So you can see that what they're going for is really that military use case. And I think for this type of technology, that makes a ton of sense. Some of the use cases like streaming and object recognition, they would be able to do in darkness. I think is really, really good. And love how they set up the industry problem. Which is basically like you have these old technologies that can't do night vision really well, and that are really expensive.

They're going to bring down the price point on that using digital, you really need specialists to do this, and that's what they are. Absolutely love this business. Thomas and Lucas, congratulations. Let's move on.

Up next. Yondu. Robotic foundation models for truly intelligent control. Very interesting. Let's jump into this.

Robots are not in nearly as many places as they should be. It's getting difficult to get them to do all the tasks that we want them to do. And they lack the ability to interact with us. To remedy these issues Yondu is creating more readily accessible robotic intelligence. And we're targeting mobile robotics first.

Makes a lot of sense. I like this. I like this a lot.

So what's the problem here. Robots are widely underutilized in multiple industries. Why? They have limited ability to interact with humans and it's difficult to use them. as it is. We either need developers to create custom software or skilled operators to control robots in order to carry out specific tasks. Okay.

So let me pause there really quickly, because I've talked to a lot of investors about this market and many people believe that robotics is actually the best application of artificial intelligence. That it's going to be much bigger than the chat applications that we're seeing today. The idea here is that in industries like industrial robotics, military transportation, that we're going to see all of these applications of artificial intelligence where robots can actually do way more than they can do today. Because they're now intelligent. And. The thought goes. OpenAI ChatGPT They've already won the chat market. The LLM market is inundated. So now we need applications. And the applications are going to happen in robotics. So this is on the back of a trend that I really believe in. There's some fascinating companies in the world of artificially intelligent robots. Companies like Figure AI, which is building autonomous robots. Tesla has it's robot. people have probably seen some of the work that DeepMind did back in the day. So, this is pretty interesting.

I worry about this idea of robots being underutilized in multiple industries, because it tells me that they're not going after a specific industry yet. I think for this type of technology, That's actually okay.

There are enough great robotic and AI companies out there that if they can figure this out, somebody would want to snap them up. But I think it really helps when you drill all the way down, especially in robotics where it's difficult, because when you say multiple industries, You can be talking about anything. Right. So. I think some clarity would help there, but maybe they'll get to that a little bit later in the pitch.

Okay. So we have a demo for law enforcement. Let's see what we got here.

Hi, Our company is called Yondu. We're about to do a demonstration of some of the capabilities. of our Voice controlled drone. Take off, fly up 0. 5 meters turn around and follow the guy wearing the orange hoodie. Alright, So here, you're about to see the drone follow just given natural language instructions.



Is interesting.

Yeah, I like it. I like it. I think that's really interesting. Basically, what we saw was exactly what he described.

This is why I was saying if they just figure out the technology, there are going to be companies that are really excited about this. So let's see some of the applications that they've mentioned.

Drones for firefighters and photographers. Robot dogs for construction and companionship, hospitality bots in hospitals and homes.

So the idea is that these mobile robots are going to be the first wave of robots that become mass market using artificial intelligence. I think that's a really interesting pitch.

So let's figure out who the team is here. The founders met while studying at MIT and have been involved in various escapades throughout the years. That's an interesting way to put that. From hacking hi-jinks to grueling late night projects, the two have done it all. And we've got a photo of them with a mattress on the subway.

Well, that might be even more impressive than the robots that they built and they got a mattress into a subway. Incredible well done team.

Michael was studying robotics at MIT and has worked on building an open source bipedal robot and an autonomous train track cleaner. Very interesting. Tahmid studied computer science and Aero Astro at MIT. And was a graduate student researching uncertainty, quantification for deep learning.

P asking someone start a same day furniture delivery company. That's in reference to them being on the subway. That's funny.

So I really like what this team is doing. I think contrasting to what I thought in the beginning, they actually are going after a few very targeted use cases, which makes sense. They're going to sell this software to robotics manufacturers. And as those robotics manufacturers proliferate, they will expand the number of use cases that they're doing. I just think what happened early on in the pitch was they focused more on broad applications. I think there is a good time to just drop all your specifics, but they have the specifics. Well done, Michael, and Tahmid Yondu I really liked this company.

Moving on.

Kontractify. Painless RFPs. Powered by LLMs. Okay.

Hello, everyone. We're Justin, Anand, and Aiden childhood friends and co-founders of Kontractify. Kontractify uses LLMs to help enterprises run RFPs.

So RFPs are a big part of how large companies will contract for software and other services that they get. RFPs are requests for proposals. And this is actually a very, very big industry.

Okay. The problem. Enterprises buy large amounts of goods and services through RFPs, competitive bid processes, where they invite multiple seller slash service providers to participate. These buyers launch RFPs on extremely outdated platforms. Often costing hundreds of thousands of dollars. And have UI slash UX stuck in the nineties. Buyers get so frustrated with these tools, they ignore them altogether. And revert to Excel spreadsheets and email. On the other side, sellers who receive RFP notifications, hate filling out responses on these platforms. Leading many to not participate.

So I've gone through an RFP process before several of them actually. And this all rings true to me. RFPs are an absolute nightmare when you're dealing with big legacy companies. And the challenge when you're a startup, is that some of those big companies can make or break your business. Like, if you get the deal with them, you're going to win. If you don't get the deal with them, you might die because it takes so much time to get down in the process of doing it, that you don't have time to chase other customers. So. Love the problem.

And they have a little GIF with a demo here. Looks pretty slick.

I like it.

So let's learn about the team. We've been friends for 18 years and have deep engineering and ops experience at companies like Palantir, Datadog, Affirm, and NextDoor. And we've got a nice photo of them standing outside the YC logo. I like my guy in the shades on the right. Very slick.

Yeah. Good pitch. Good business. Curious about what the traction is like, how the uptick is coming. The challenge for this is that the only way for this business to work is if they get large companies to buy into it. So they themselves are subject likely to RFP processes to get this software in the door. So, Hey, you know what? Dog food, the product that might be a really interesting sale to make, but I like the problem.

I like the team. Well done guys.

Up next. Base pilot. Your AI employee to automate repetitive browser work in minutes. Okay.

Basepilot. It creates AI assistants to automate repetitive and manual work across a browser. We help companies save 30% of their time weekly and cut costs, freeing up resources for more high value tasks. For example, finance, recruitment, or sales teams can automate mundane back office tasks within a few clicks and a quick demonstration.

I think this is good, the thing that a system like this is going to run up against is that there are vertical software as a service providers, SaaS providers. Who have interfaces that all of these people already live in. And the trend that we've seen in AI is that these systems of record, which vertical professionals are using.

So things like salespeople, recruiters, Financial services professionals. These are the systems where they spend most of their days. And those systems all now are trying to implement AI. So as a startup, you're in a race against the system that already has all their data and all of the stuff that you would probably need to get people to use it. However, a lot of these companies, even though they're tech companies cannot move as quickly as startups can.

And so that's really where the opportunity lies. It is you can move much faster than those guys can. You can probably wow a few people, but you might have to steal those customers before the incumbent systems find a way to implement this into the workflows that they already have.

Another outcome for some of these companies, is that the incumbent systems will just buy them out. But you can imagine that let's just take the example of sales. Salesforce is spending billions of dollars trying to get AI built into it systems. Do I trust Salesforce to pull that off? No. But I don't think salespeople are going to stop using Salesforce. Right? And I think over time, what you'll see is integrations built into these systems of record that actually do a lot of the work.

I think that's interesting what they're doing. It's very competitive. But. Oh my God, but I like it. Sorry. They have a GIF in here of a guy throwing a computer, which again, that gets major points for me. I love computer throwers. All right. So it looks like they have a demo here.

BasePilot. com

BasePilot helps companies build their own AI assistants to automate repetitive back office work.

For example, sales, recruitment, and finance teams can automate boring admin tasks. With a few clicks and a quick demonstration.

So let's jump into it and see a quick example of how BasePilot really gets to it.

Over here. We have a list of companies in sales navigator. What we will be doing is taking these lists of companies, adding them to our HubSpot Going ahead and finding the CEO of each of these companies back in sales navigator, and also adding them as leads into HubSpot.

So let's see how BasePilot does this task. On the right. here, We have BasePilot and a list of workflows for this one it's called company research. Let's go into it. And we see a bunch of steps and we can start the workflow. What basePilot is doing now is completing these tasks and taking over the UI to do so. So, what we're doing here is BasePilot has gone ahead. Got Airbnb added it to HubSpot. It's now going to go and look through the Airbnb employee list. to try to

Okay, so that is cool. I think if I were them. I would go all in on integrations. I would find a way to get this into the systems that people are already using. And sell those companies hard, build loyalty. And then try to pull them out of those systems into my own system. That would be my strategy for trying to build a business.

It's going to be difficult if you are subordinate to some giant company, like, let's just say, they're in the Salesforce marketplace. Then Salesforce decides to build this themselves and they start throttling them that would just cause all kinds of problems for their business. So I think it's smart for them to build into systems that already exist, and then turn those users into independent customers on base pilot's own systems. But I really like this.

So a little bit about the team. Pascal and Ken both independently traveled to San Francisco in the summer of 2023 from Germany and New Zealand, respectively, who I love internationals. We came here without knowing each other to explore the startup space and build the future.

While in SF, we met through common friends, vibed well and started working together, you know what. Anyone who has the right vibes, I'm on board with that. Vibes are so important in startups and these guys are letting you know, they got good vibes. They look like they got good vibes. I love it. Congratulations on launching. This is cool technology. Best of luck to you guys. I think what you're doing is really cool.

Okay. Up next . Sevn. AI graphic designer. Sevn designs, ads, presentations, and UIs 10 times faster than humans. There are not enough graphic designers in the world today. Businesses need graphic designs across a lot of critical use cases. For example, to create ads, to promote themselves. Beautiful landing pages. User-friendly UIs. Presentations to their clients. However many businesses can not afford to hire a designer.

Okay. So I'm going to say something real quick about that. Many tiny businesses cannot afford to hire a designer. So that tells me that their use case primarily is going to be small businesses. Small startups. Even those, I think. Designers. You can find designers. Not necessarily good designers. You can find designers for pretty cheap, but this is exactly the sort of thing I would expect. Small companies to use.

So. I like that.

Okay. So we've got an AI that can design they've got a few GIFs describing exactly how it does it. Essentially it takes a prompt, it gives you some options, allows you to make edits as you want. But the copilot agent will do some of the most difficult tasks in design.

And this is exactly what they say.

Professionals use Sevn to create designs for their work needs. They talk to Sevn as they would talk to a designer to iterate over designs until satisfied. Sevn remains tireless.

I'm actually really glad that they mentioned that point because I've worked with a lot of designers over the years. Design was actually one of the things that I personally spent a lot of time learning because of frustrations I had with designers. There's a couple of cases that I would highlight here. There's one case where you've hired an outsourced designer. So somebody who you're just going to pay to do a project, but they're not part of your team. They're an independent contractor. Those designers effectively are going to charge you every time they talk to you, that's in effect what's happening.

I mean, you might pay them a lump sum across the total amount of the project, but they're only going to give you a certain number of revisions. They're going to complain about lots of things that you ask them to change in it. And it ends up being really expensive. Not necessarily more expensive than a full-time employee, who is a good designer, but it does cost real money to get designers to do stuff. So that's the outsource case. The case of the designer on your team, is that depending on what their background is, A lot of them can have very unique and complex processes that if you, as a business person have a vision, Their design process is going to take your vision and turn it into something different.

Sometimes they're right. Sometimes they're not. I think at the startup phase, everyone is just in discovery mode and it helps to have a cook who's in charge of the kitchen. That should probably be the CEO or the CTO. Designers rarely are the ones who fill that role, although sometimes they are, but you're just going to learn through iteration in those cases.

And so I think having a tireless AI designer who can go back and forth with you is valuable. In both cases, I actually really liked the way that they framed that.

Made with love by Jean and Srijan. We have spent the last decade working in AI. John's PhD is on LLMs and John has worked on large scale ML at Meta and geo Sevn. Both of us have had to spend hundreds of hours creating designs for our research projects. And work. We realized we know a great design when we see one, but we don't know how to create one. Okay. I like it. I like it.

I expect. Just like with the last startup, we looked at base pilot that what these guys are going to have to do is find a way to get into existing design suites. Like when I do design, I'm spread across a number of different suites. Many of which are very forward-looking Figma is the one that jumps out to me the most.

But Canva is as well. What you run into is that these systems are very much trying to do exactly what Sevn is trying to do. And just like in the case of Basepilot and sales, the system of record is going to win in those cases, unless you can find a way to integrate with it, get them familiar with what you're doing, and then try to pull them off platform in the long run.

They'll pay you a lot more. So I think again, that's the same play, but what they're doing is really smart and. I like what I've seen. Good job Sevn.

Up next. Yoneda Labs foundation model for chemical reactions. Enabling new chemistries with AI. Now this just gets the science nerd in me, all

hyped up. I love this. Yoneda Labs is building the foundation model for chemical reactions. When chemists need to make a drug, we use AI to define parameters such as temperature, concentration, and catalyst. This enables new chemistries and makes synthesis faster and cheaper.

And so.

Oh, man. That was great. That was actually really great. So there's a photo of them. They're both standing outside the YC logo, but then we have a third co-founder who's laying down on the ground in front of it. Spread out, across. I love this picture. Good job guys. We are three university of Cambridge grads combining expertise in computer science and chemistry with a leader in the field.

Professor Scott E Denmark as our advisor. Combined. We have experienced working in Bayer, Berkley AI research, Jane Street, Optiver and Cisco.

Jan ranked top 11 in the international chemistry Olympiad for three years in a row. Wow. Competitive chemistry. Well we saw MathDash, the competitive math startup. I guess there's competitions in every field of math and sciences. I love it. So what's the problem. There are more than 10 million different compounds created each year. Chemists need to figure out how to make them. To do so they often rely on literature, data.

Unfortunately, this data almost never contains the exact recipe. And the ones out there are often misreported and irreproducible in a lab. This means that chemists still need to frequently rely on their intuition. And when that fails, a potentially life-saving drug might never see the light of day. Very very interesting.

So they're building a foundation model for chemical reactions. We're generating our own dataset of chemical experiments. We will use it to train a powerful model, capable of defining good conditions for any organic reactions.

So I'm going to pause there for a second.

Training data for large language models is the new currency in AI.

Everybody talks about compute. That's right. You need data to train a model, to get the model, to do the things that you want it to do. By focusing on an area where you definitely need specialists and creating their own dataset that gives them what many VCs would call and moat. Their data is their moat.

Other companies can't train their own models, using the data that these guys have put together. That means that they are insulated. If their model is the best model on the market, those other companies would have to replicate their data set through their own experiments in order to do that. So I think the argument here is going to be. Let's invest a huge amount of money creating this amazing data set for this niche in chemical reactions. Then with that moat, let's go and win the entire market.

So anybody else who wants to come in is now going to have to find a way to reproduce that data.

Like it.

We've performed feasibility studies on smaller datasets and confirmed our beliefs that if we choose the experiments very carefully, we will be able to cover three of the most popular reactions in organic chemistry with only 20,000 data points. That feels like a little bit less of a moat it's possible to copy it with that few data points, but still, first to market matters a lot.

We're seeing that with open AI and chat GPT. We've already created an app that learns from the chemist experimental data as they run their reactions in the lab. This allows them to find the optimal conditions in days instead of weeks or months. That's great.

So I really liked this. This is exactly the kind of business you look for as a venture capitalist. I love the way they're approaching it. I love the team. I think this is going to be very successful. Congratulations guys on launching.

Up next. X pay. Go global effortlessly. Our single API helps businesses in India and SEA easily accept international payments in 50 plus markets via local payment methods and cards. All right, let's jump into this one.

Who are we. This is Aniket and Rakshit. The team behind xPay. For a long time receiving international payments has been a pain for online businesses. Our goal is to abstract out this pain. So that receiving an international payment feels as easy as receiving a domestic one. And we've got this nice photo of them on a black background.

So one of the few who are not outside the YC logo and they've got their t-shirts on, which love the smiles. Very charismatic. We have faced this problem firsthand in our previous businesses, serving international customers. And so we're building a single hassle-free API to help businesses in India and Southeast Asia accept international payments. We launched our beta two weeks back and already have an oversubscribed waitlist of 200 plus companies. As we try to keep up with the demand. Here's a video talking about our mission. All right.

Let's see what we got.

Hi. if you're probably watching this video, you have some idea of xPay. but you must be wondering what is xPay? Well, I'm Aniket, The founder and CEO of xpaycheckout. com. And at xPay we're solving the problem of international payments.

Businesses in india and Southeast Asia very often try To go to international markets to make, to increase their revenue learn more from their customers and have a bigger footprint.

However, Going international is not as easy. When you start to explore the solutions, you realize none of them work perfectly. You realize that you have to stitch together many different solutions to go to UK, Africa, or Australia.

Okay. I think some of the people listening might be thinking. Wait, doesn't Stripe do this? Can't you do this with like Braintree, PayPal or whoever else. The short answer is in a lot of cases, no. So a few years back there were a wave of companies. And Stripe actually bought several of them who were saying exactly this. We're going to take the payment rails international. Payments. Cross-border international payments, still very, very difficult for companies to do. Stripe isn't in every country. I don't know exactly all the countries that they're in, but if these guys are saying that Southeast Asia and India are underserved and they have a way to get in. I think they should absolutely do that, but it sounds to me like they're essentially building Stripe for India and Southeast Asia.

We have a little graph here where they're saying effectively they can integrate businesses in India, UK and the USA with these other third party providers.

xPay is a single API that unifies your integrations. It provides you with all major gateways in the world, like Stripe and razor pay pre-integrated. So this is essentially like a productivity play for payments. So they're saying. Hey, instead of having to integrate with like 6 different international payments providers to take your business fully global. You can just integrate with xPay and we'll connect you to all of these other providers. But specifically, I think they're making a claim about being able to sell southeast Asia and India as a target market, as opposed to trying to sell companies here.

Cause I think a lot of American companies, especially startups are perfectly happy to just use Stripe. And I think a lot of European startups are perfectly happy to just use Adyen. Right? So there's a reality here that the people who might be most served by this are people who are operating in markets where these large behemoths haven't yet gotten a foothold. I have to imagine those guys are sniffing around the markets that xPay is going after, but I think what they're doing is smart.

Up next. Pivot robots. Automating high mix manufacturing with AI.

Pivot robots build software to tackle the next frontier of manufacturing automation. I absolutely love how many robotic startups that we're getting here. This is cool.

We are building this. So we see a robotic arm that is picking up tools, helping them with the manufacturing process and moving them along.

That's cool. We are Ex-Google and Meta robotics engineers that are using foundation vision models to solve the riddle of high mix to redefine American manufacturing. American manufacturing. Yeah.

Love that. Us manufacturing is making a comeback for the first time, since the 1970s. But there's a problem. Nobody is filling these jobs. Decades of decline and entire generations shying away from manufacturing jobs have left the industry facing and expected 2.4 million jobs going unfulfilled by 2030. The obvious solution is to use robots. But 75% of all US manufacturers are effectively unable to use traditional robotic automation due to their high mix production. They produce a wide array of products in small quantities.

Traditional automation is rigid and needs new fixturing and reprogramming for each part. And doesn't provide ROI for high mix manufacturers since so little of each part is produced.

Okay. So I'm going to pick a small nit with something that they said there. That no one is filling these jobs. And the obvious solution is to use robots. I think there's another solution where you could train people to do the things you want them to do.

But this high mix issue is a good point. The robots might be more versatile than people. trying to avoid the claim that they are killing jobs for Americans, right? They're very concerned about that. I can see that. So they're just saying, Hey, there's a shortfall here. There's no way to fix it.

But yeah. I got it. Good move.

If America is unable to capitalize on the current manufacturing, AI, boom, we lose out on $1 trillion per year in opportunity costs. And the chance to strengthen our weakened industrial base. High-mix manufacturers are a crucial component for this. And they urgently need adaptable automation solutions that can cheaply handle their large variety of parts.

They keep talking about the interests of the United States and of America. It's kind of a nationalist argument that they're making and they're straddling a few sides of that because I think most politicians would say we don't want to lose American jobs.

Like if we bring back manufacturing, what we really want is Americans to fill those jobs. We don't necessarily want a piece of hardware to take all of the jobs away from the people who might be doing those jobs. But strategically, you also don't want to have all of manufacturing live in countries like China, India, Taiwan.

So you want to bring some of that back on shore. Our challenge ends up being that building American factories is very, very expensive. It's way more expensive here than it is over there. So we're in a bit of a conundrum. I do think that robots will be a huge part of the manufacturing push here. Simply because the cost of employees and unions and everything else with managing a population of factory workers is very, very expensive.

And so companies kind of just look at the bottom line. They say, well, yeah, I mean this local government or state or national government federal government might give me some subsidies, but it's still cheaper for me to make my stuff in China. So until that number crosses over in the other direction, I'm going to keep doing Chinese manufacturing and keep outsourcing it.

No, one's going to see what happens in those factories or the fact that they pay people. You know, almost nothing and make them sleep on the factory floor. So. They basically go ahead and say, yeah, we'll hire some of the biggest Chinese companies to do this for us. I think the move to bring this back onshore is a good move.

So I like it.

Our solution. Pivot robots combines off the shelf, robots and vision sensors with our pioneering AI vision control software to give industrial robots the ability to adapt. Our first product Proteus addresses the dangerous and unpopular task of grinding metal parts. Offering zero shot adaptation to new parts with the same position and efficiency. Well, I really like this.

We are currently engaged in a pilot to deploy 10 Proteus systems at large cast iron Foundry that produces over 200 parts a year. And we've got a photo of them here with their robot. I really liked this a lot. So let's talk about the team. Siddharth has worked on self-driving cars at Uber ATG, AI at Meta, and most recently was designing multi-agent algorithms for warehouse robots at nimble robotics. Vignesh optimized motion planning for general purpose robots at Google X, researched LLM applications for robotics at the Atkinson lab, and worked on manipulation at ABB. Both of us are inspired to use our robotics experience to solve clear, real-world labor problems after watching our respective parents careers in those industries. Agriculture and manufacturing.

This is cool. This is really cool. Like what the other robotics companies, I think it's sitting on, the back of a big trend, which is people seeing artificial intelligence. this case, computer vision also as the future of robotics. And I think what they're doing is really, really smart. I like it. And I also like how they called out their traction.

Guys, congratulations. Great work.

Up next. Ubicloud. Open source alternative to AWS.

Ubicloud provides IaaS that's infrastructure as a service cloud features on bare metal providers such as Hetzner, Leaseweb or AWS bare metal. You can self host our software or use our managed service to reduce your cloud costs by three to 10 X.

There's a lot in what they just said. Put aside, the jargon involved. What's happening here is you have companies that have to spend huge amounts of money on the cloud.

So think just storing your data on the servers of some large company like AWS. This cost is often one of the biggest costs outside of employees that startups will incur. So I think it's really smart to basically say, Hey, we can cut your costs by a lot.

All right. So let's hear about the team here. We're Umur, Daniel and Ozgun from Ubicloud. We're on a mission to rebuild the cloud market with different values.

Oh my God. I wish they'd put this picture at the very top. This should be the lead picture. So, what we have here is these guys in YC's San Francisco office, which YC's colors are orange, primarily. It's like on the logo, it's just this bright orange and they're all wearing orange jumpsuits. So they look like prisoners. That is hilarious.

And they've got a quote from Han solo here. Never tell me the odds.

This is great. I love these guys. I don't even know if we need to go through the rest of this pitch, like they just completely won me over.

So these guys are actually former YC founders, successful YC founders. Incredibly successful team. So Umur five years at startups then co-founded and led Citus Data to a Microsoft acquisition. He led product teams within Azure. And he was a YC visiting partner for the previous two batches. Ozgun spent four years at Amazon then co-founded and led Citus Data to its Microsoft acquisition and also led teams within Azure.

Daniel was on the team at Heroku. He amazed everyone with his skills in building managed cloud services at Citus Data, Azure, and crunchy bridge.

Okay. So let's talk about the problem here. AWS, Azure and Google cloud make life easier for startups and enterprises, but they are closed source. You have to rent your computers at a huge premium. And they lock you in, it gets worse after your credits expire. Although the cloud came in with savings in its early years, the market evolved into a few players that overcharge for non-differentiated services with increasingly complex products for standard essential services. Each vendor optimizing for their walled gardens, leaves customers stuck maintaining their apps and infrastructure across incompatible clouds.

The cloud is here to stay. But we believe a simpler, secure, and open source version of it will have a lasting impact.

That's great. I really like this. I think it's an incredible team. This is the kind of team that you just close your eyes and invest. I mean, I think it's all a really smart explanation. Like, I love the way that they set it up and the right. But they've been so successful in the things that they've done before that I would just bet on them almost regardless of what they're saying.

But I really love the idea. Bringing down cloud costs would help thousands and thousands and thousands of companies. And I love the concept. Congratulations guys. Moving on.

Shiboleth. Automating consumer lending compliance. Enabling lending compliance teams to audit customer service chats with AI.

Lending companies use us to flag compliance violations for customer interactions and create the first drafts of the reports for the government.

This is exactly the type of boring, unknown problem that ends up becoming a multi-billion dollar business. Like seriously. It's the things that no one would think of that are actually so simple that you're like, wait. Like, how is this even a thing? But in reality, they're doing enormous amounts of work for extremely wealthy customers.

And that means that you end up being really rich. So. I liked the premise here as a business.

Hello, everyone. We are Esty and Bivu. We are building Shiboleth. I hope I'm not butchering that pronunciation. If you saw a recent Garry Tan tweet about Bookface launch live, that was us. We were chosen as one of the top products based on votes from batch and YC alumni.

Esty previously worked in the intersection of compliance and conversational intelligence. And Bivu has a decade long experience in NLP and AI. it's not quite clear what it is that they were doing or who they were working for.

I think that would end up being important. I know they have their links here, so you could probably see that, but we're just going to keep on moving. Got a nice photo of them hanging out outside of the Y Combinator mountain view office.

The problem.

Consumer lending is very litigious and consequently it is heavily regulated agencies such as CFPB, FDIC and OCC impose a lot of intentionally ambiguous regulations. Often there's just enough ambiguity that the calls and the communication warrant manual reviews. This is problematic because an internal audit system might not suffice for compliance purposes.

All right. How does it work?

Shiboleth automatically flags compliance violations in customer interactions. It does this by syncing with the company's runbooks using advanced LLMs and. NLP to parse relevant regulations.

check out the demo below. All right,

Introducing Shiboleth, The cutting edge solution for automating QA and compliance reviews in consumer lending.

Here's how it works.

Shiboleth's AI

diligently audits, customer conversations. Such as these phone calls,


compliance violations.


dive into a flagged call.


AI has identified a salesperson pressuring a buyer into a commercial loan.


AI identifies similar violations in your data. And evaluates their legal risk based on the latest supervisory insights.

Responding swiftly Shiboleth notifies key stakeholders in real time. and suggests appropriate mitigation steps.

In this example, let's send an email to the chief risk officer and the head of QA using. the built in


I like it. I like it. I don't know how strong of a moat is here because the regulations, complaints and so on, they mentioned, I have to imagine that all of that's public so anyone who wants to train their own LLM to do this can probably get the data they need and do it so there's not really a strong data moat here. But there is a big customer moat.

Getting these large financial services companies to adopt something like this generally means they're not going to pull it out very quickly. Right? Like once they adopt it, they're going to hold onto it, especially if it's actually helping them and saving them money.

So let's see about who they're targeting. Our solution is designed for financial institutions, including banks, FinTech companies. And non-bank lenders that are looking to streamline their lending compliance processes and reduce costs.

This is a space where again, it's a boring space and I'm not saying that as an insult, I'm actually saying that as kind of a compliment because these are the businesses that people don't know about, and it means that there's not going to be some line of a thousand startups that are trying to go into it. Most of the guys who are excited about AI are going to build things that seemed cool, that they could tell to their friends and family, Hey, I'm doing this and it's interesting. When you do a business like this, that's really like a back office function and it's dealing with like kind of dry, legal regulation stuff. It's not as sexy, but these are the kinds of businesses that make insane amounts of money.

So I like it. Congratulations to the team on launching. And best of luck.

All right. That's another 10 startups in the books. Loved pretty much all of them. It was a bunch of really interesting companies. I got to tell you, the winter 24 batch is really impressing me across the board. I've seen a lot of people talking about it and having talked to a number of the founders myself. It is incredible.

I don't know if I would have gotten into this batch, frankly, so well done to the teams. Thanks for watching. Thanks for listening. And we'll talk to you again soon.

Watch next