Salient: The Fintech Startup Processing $1B+ in Loans with AI

1. 0:00 – 0:22

   Intro

   1. DHDiana Hu0:00

      [upbeat music] So today I'm super excited to have Salient here, who are announcing a massive Series A led by a6c for $60 million. Congratulations. Tell us a bit about what Salient does.

2. 0:22 – 1:03

   What Salient Does

   1. AMAri Malik0:22

      So Salient is an AI loan servicing platform for consumer lenders. Uh, we started off with auto loans, but now we help manage credit cards, mortgages, anything that a consumer can take a loan for. So we work with non-bank lenders like Westlake Financial, Exeter Finance, American Credit Acceptance, as well as three of the largest publicly listed banks in the US.

   2. DHDiana Hu0:44

      And you guys went through YC not too long ago. It was just two years ago.

   3. AMAri Malik0:51

      [laughs]

   4. DHDiana Hu0:51

      This is a crazy ride to get there, so working with the largest public banks in the US, that's wild. Where did you guys get started? How did you find this idea?

3. 1:03 – 1:26

   Early Beginnings at Tesla

   1. AMAri Malik1:03

      Um, so it all kind of started from my time at Tesla. So at Tesla, we were giving out loans to help drive new car sales. And what we found was that the cost of managing these loans was very, very high, despite the fact that we were giving loans to people with great credit scores, all in California. And when we kind of dove into what is it that's really driving the spend,

4. 1:26 – 1:51

   Leveraging AI for Loan Management

   1. AMAri Malik1:26

      I think Mukund realized that, you know, we could actually automate a lot of the stuff using AI.

   2. MTMukund Tibrewala1:31

      So I think we're very fortunate with the timing on two crucial aspects. The first one was the advent of closed source models like GPT-3.5 coming out. That led us to, like, pave the way for really, really high quality demos. The second big leap came from when it came time to scale. We realized that the advent of open source models that were available for fine-tuning, such as

5. 1:51 – 2:15

   Scaling with Open Source Models

   1. MTMukund Tibrewala1:51

      the Llama series, really, really helped us scale from, like, a few hundred dollars per day to over $100,000 per day. Um, along with that came vLLM, which was the open source inference library that really, really allowed anyone like us to be able to grab one of these Llama models, serve it up to millions of requests per minute.

   2. DHDiana Hu2:08

      You were one of the first companies to really take advantage of the alternative with open source

6. 2:15 – 2:39

   The Impact of LLAMA 2

   1. DHDiana Hu2:15

      big models. So Llama 2, that was released in July 2023, was the moment for you. Tell us about that. What, what, what was that... specific about the model that made it possible?

   2. MTMukund Tibrewala2:26

      So I feel like that was a really big week for us when that model came out, because for us it paved that big gap between having a high quality demo, but still only doing like 100 calls a day, to being able to sustainably scale to hundreds of thousands of calls

7. 2:39 – 3:12

   The Magic Demo and First Big Customer

   1. MTMukund Tibrewala2:39

      per day.

   2. DHDiana Hu2:40

      I remember when you showed me a demo. I got this call from, uh, the voice of, uh, Steve Jobs.

   3. AMAri Malik2:46

      [laughs]

   4. DHDiana Hu2:47

      Uh, it was very convincing. This is starting to work. AI has passed this moment of the Turing test, and I was, like, impressed. And you got that working around that summer. And with that demo is how you got one of your first big customers. Tell us about that journey around... That was in the summer, fall of 2023?

8. 3:12 – 5:02

   Cold Emails and Westlake Financial

   1. AMAri Malik3:12

      Yeah. Um, so I think we cold emailed every single auto lender under the sun.

   2. DHDiana Hu3:16

      [laughs]

   3. AMAri Malik3:17

      And-

   4. DHDiana Hu3:17

      How many emails were you sending?

   5. AMAri Malik3:19

      Like, 500 a day. I was just, like, sitting... Like, I was... I remember being in a room just looking at a laptop from 8:00 AM to 9:00 PM just sending emails. [laughs] And, um, I think this is probably the best thing that, that ever happened to the company was, uh, getting a response from Westlake Financial. I mean, they're a massive auto lender in the US, you know, more than $25 billion of assets, and they were down to meet with us. And, you know, at that time it was just me, Mukund, and Bri working out of my bedroom in San Francisco. And here was this massive company that I think saw the potential of this technology and knew that we were the people that could actually do it. And so while they were thinking about other vendors, what we ended up doing was just moving there. Like, we packed up our bags, we left SF, we moved right next to their office, and we spent, I mean, a year and a half really getting them live. You know, and I think it's thanks to their embracing technology, that they wanna take risks, they wanna be at the cutting edge, and they're willing to partner with us, that we are allowed to really build, I think, cutting edge solutions. Um, and so thanks, you know, we've now done more than a billion dollars in processed transactions in the US. We've interacted with more than 3 million unique US borrowers. We dial more than 400,000 times per day. And I think it all, all came from that cold email. [laughs]

   6. DHDiana Hu4:39

      Wow. That cold email and that Steve Job call demo-

   7. AMAri Malik4:43

      [laughs]

   8. MTMukund Tibrewala4:43

      [laughs]

   9. DHDiana Hu4:43

      ... was what it took. I think that was definitely one of those magic demos that you, you gotta, like, use this. I, I was, I was so impressed. You guys got to a couple million in revenue a year with just the two of you.

   10. AMAri Malik4:59

       [laughs]

   11. MTMukund Tibrewala4:59

       Mm-hmm.

   12. DHDiana Hu5:00

       How did that happen?

9. 5:02 – 5:35

   Relentless Customer Focus

   1. AMAri Malik5:02

      I think it's honestly just being, like, relentlessly customer focused. Like, we had these amazing tailwinds of AI getting better and better every month, but I think the thing that, you know, Mukund does really well, and I think I do really well, is we weren't just in the lab. Like, we were with our customers every single day. We were plotting out all their workflows. And when it came to actually deploying the product, we didn't just say, "Here it is. Go figure it out." It's like, "Here it is. We'll put it into production with you." And I think that kind of like relentless, like, commitment to customer success allowed us to scale very, very quickly.

10. 5:35 – 6:19

    Forward Deployed Engineer Playbook

    1. DHDiana Hu5:35

       I think the thing that was very hardcore, just double-clicking on what you said, is you moved from San Francisco to live next to your customer [laughs] until they were successful. You were effectively deploying the forward deployed engineer playbook that Palantir gutted down to get this massive enterprise live. And in that process is, what we've seen at, at YC is a lot of these AI companies like yourself that are growing very quickly, is what it takes to go live and quickly grow. So you went from then-... couple million. A year passed, and y- you got to tens of millions in revenue, and the team was only 10 engineers?

11. 6:19 – 7:13

    Scaling with a Small Team

    1. DHDiana Hu6:19

       How do you guys make that happen?

    2. AMAri Malik6:21

       Honestly, it's due to the fact that how great our first 10 hires were. We hired XYZ founders, we hired really young, crack talent, and we hired people who wanna move really, really quickly. And I think what we're learning in this current, like, in our experience with, like, deploying AI solutions, is actually it- it's not the size of the team that really matters. It's, like, how committed is your engineer to getting a working product out? And having that, I think, like, relentless push from every single one of our engineers, who makes us better, they make our product better, they make our customers happier, I think is, like... It's all thanks to them.

    3. MTMukund Tibrewala6:55

       I think one thing we realized early on was that we are gonna make every one of our engineers responsible for a major customer. They were not used to that coming in from a couple of years at Google. They became the account manager, they became the forward deployed engineer. They were talking directly to the CFO and to the VP of servicing, and they were completely unprepared for that. But, like, three weeks in, it was like they've been doing it

12. 7:13 – 7:47

    Hiring for Growth

    1. MTMukund Tibrewala7:13

       every day.

    2. DHDiana Hu7:14

       So what are the kinds of people that you're hiring now with all this money that you raised?

    3. AMAri Malik7:17

       Yeah, so I think, like, now that we've raised this money, we've had so much pent-up demand in the market for so long. Like, the biggest constraint in our growth has just been ourselves, and where we wanna go with this is hiring really, really high agency engineers, high agency product managers, salespeople, to help take us no longer from zero to one, but from one to 1,000.

    4. DHDiana Hu7:38

       I think voice AI became a very good foot in the door for a lot of your customers, but now you ended up doing way more.

13. 7:47 – 8:24

    Challenges in Scaling Voice AI

    1. MTMukund Tibrewala7:47

       I think one of the challenges that we had not yet anticipated as we were trying to grow was that it wasn't just about having high-quality voice AI. It wasn't having, like, low latency, good interrupt behavior. It was about scaling to hundreds of thousands of calls in a safe manner according to US regulations. It's about making sure that we only dial an account when it is legal to do so, and it is ensuring that our system is able to, like, simultaneously manage thousands of dials we do simultaneously.

    2. DHDiana Hu8:16

       And as part of that, you figure out a lot of really hard technical pieces. You guys got to PCI compliance in record time.

14. 8:24 – 8:56

    Navigating Regulations and Compliance

    1. AMAri Malik8:24

       Yeah, I think the big thing is, like, we work in such a regulated industry. We have laws at the federal level, we have laws at the state level, we have s- you know, in some cases, [laughs] even laws at the district level. And so I think a bulk of our time in building the solution is, how can we do this in a way that protects the lender, but also makes the consumer feel safe? So it's building in safeguards on bankruptcy protections, on legal representation, on codifying TCPA. So, you know, I think we've now become big experts [laughs] in the US consumer lending space.

    2. DHDiana Hu8:54

       So what does the future look like for you?

15. 8:56 – 9:43

    Future Vision for Salient

    1. AMAri Malik8:56

       Our goal is to be the system of record for every loan in America. When a lender books a loan, it should come onto our software, and that should be touchless for that lender from start to finish, from the date of origination to the date of charge-off, and that means us building a CRM. It means us building an accounting system, and that means us building workflow automation tools. It means us building an, an AI contact center. And the end vision of what Salient does is, like, we should lower the cost of credit for the American consumer. Like, those costs should no longer be blindly passed from a, like, you know, a lender to a borrower. We should make that spread almost zero, and so that's what we're, uh, trying to accomplish.

    2. DHDiana Hu9:36

       Thank you so much, Ari and Mukund, for coming, and congratulations on your massive Series A.

    3. AMAri Malik9:41

       Thank you so much for having us, Diana.

    4. MTMukund Tibrewala9:42

       Thank you so much. Really appreciate it, Diana.

Episode duration: 9:45