Luminai raises $38M Series B to automate health system back-office operations — hits 90% on-prem pass rate

We'll talk to you soon. Have a good rest of your day.

Cheers.

Uh up next we have Lumini and Cassava Dina Kuran in the waiting room. We'll bring him in to the TVPM Ultra Dome. They have raised a $38 million series B to scale AI automation for health systems. How are you doing? Good to see you.

What's going on?

Doing doing great. Great to see you both.

Great to see you. Um I believe it's your first time on the show even though we've met in the past. Uh please introduce yourself and the company.

Yeah. Uh I'm Quav, one of the co-founders and the CEO of uh of Lumini. Uh help uh large health systems uh drive automation in their operations and back office.

What is uh so is this a diffusion story or is this a uh like a foundational AI research problem like what what what is the secret to actually driving improved healthcare outcomes in large in these larger organizations? Yeah, I think you know the the models have kind of got to a place uh probably in the last maybe 6 to 12 months where um where some of the opportunities to actually drive automation end to end in a very reliable manner is is very possible. So it's 100% a diffusion uh story at the moment because the reality is like uh you know American healthcare still runs on faxes uh in many ways and so you're kind of at this opportunity now to to dramatically change how uh how it's done in the first place.

So G uh would love an example of like a specific back office task. Is this like passing information to insurance providers processing claims like what what are some specifics? Yeah, I'll give you a very uh concrete example. Um, so imagine like you know um one of the customers we work with is the Cleveland Clinic. Um, and uh they're obviously one of the most advanced sort of academic uh research organizations as well as hospitals in the in the world. Um, and so uh there are patients who come from all over the world to Cleveland Clinic and the way they actually get uh care is initially through uh something called a a referral. um right and and that referral basically uh gets sent from um you know small clinician practices to massive hospitals that are based in you know even all the way from the Middle East and uh they're all uh unfortunately um or key central lumini sent via a fax machine um and these are sort of documents with handwritten notes where people are sent saying hey this is Kesh he has some u unique sort of um left knee pain uh that require requires you know a specialized attention. Um here are the details and there's a huge sort of operations team on uh uh the Cleveland Clinic side where their job basically is to look at every single document read every single sort of um handwritten note. Uh and by the way these facts

so so sorry to interrupt but you guys so so just so I have it correct you guys are making humanoid robots that can write handwritten notes to fax back through a physical fax machine to the original sender. You know, I think five years ago that would have been an easier solution to solve uh in healthcare uh than than try to get it implemented. But but uh it's all all enterprise software. Uh so it's it's just literally a virtual inbox agent that uh you know triages each referral that comes in uh puts in the high-risisk patients first and then processes the referral um in an automated way. But that that's sort of an example of of many other problems that obviously exist within the the operations of the health system.

That makes sense. And so are you guys uh doing the kind of forward deployed model where you basically say, "Hey, we we have a good understanding of current AI capabilities. If you let our team into your space, we'll figure out individual workflows to automate or what's been the the the approach so far?

Yeah, I mean um healthcare is extremely nuanced. Uh you have to understand the specifics pretty pretty deeply. Um and so without actually being on the ground living in the offices of your of your customers, it's difficult to sort of scale solutions. And so you know about 20% of our team comes from Palunteer and so we've taken a very deep uh forward deployed approach uh to every every customer you work with and you know we work with some of the largest institutions in the country and so sort of it lets us afford to actually be able to to do that. How much about uh deployment diffusion is uh gated by regulatory or approvals versus um it's a less maybe you know obviously there's a lot of technative people in these health care systems but uh they might be a little bit less online a little bit less uh aware of the of the progress and so there's just an education element um versus you know understanding the cost tradeoffs and actually implementing the the the processes. I mean there definitely a a massive amount of like uh barriers to entry and being able to actually drive uh or or actually implement um uh AI systems in these places prim and for good reason right like this is like extremely extremely uh sort of sensitive healthcare data uh that's that's flowing through these systems and so there's the right amount of sort of guardrails put in place uh but once you're in uh there's a dramatic amount of work uh that's uh that can be done by software And that can be done by you know um AI in the first place and uh it's uh it's sort of a you know there's definitely areas we shouldn't touch like deep deep clinical decision- making and and threads around that that you know this is why we have doctors um but there's a huge amount over a trillion dollars just like wasted administrative uh operations work that that uh yeah today uh can be done by software systems.

How are you thinking about onrem these days um with the AI context? to mean these models are so big. The latest ones seem to run on NVL72s. That feels like a big lift in capex if you're going to stuff that in a closet somewhere or in the local uh IT cabinet. Where where do you see uh like are we going to be seeing like local inference happening for uh for regulatory reasons or is it more about uh just interfacing with on-prem systems because they exist and then this will be actually an accelerant to cloud adoption. um you know we've offered on prem to every single health system we work with really but only only you know 10 to 15% of them have actually taken us up on it and um and so that's an interesting sort of one data point but the primal reason to offer onrem in the first place is cuz uh because of the amount of data that you're sort of handling and you don't want those u that that data to basically uh leave your leave your premises in the first place and so uh we've chosen to go go onrem with with you know even certain workflows uh with with large health systems. Um but but funnily enough uh we thought it was going to be a massive um need u when in reality uh it's it's definitely a check box that they do to make sure that you can actually do it versus actually deploy some of this in in in full production.

Interesting. I mean um how how dramatically is uh is this the the the the quantity of data produced in health care environments growing right now? We looked at this one company that was uh just a an audio recorder that would transcribe everything that uh someone would say. We've seen a bunch of these targeted in in tech audiences, but this one was specifically for doctors to take notes and then have a running transcript of everything that they said as opposed to needing to scribble a bunch of notes. it feels like we could be at uh some sort of like data production uh inflection point but what are you actually seeing in terms of the quantity of data that's being produced in the healthare system?

Yeah, I mean there there sort of obviously a variety of folks who've done uh research on this but I think the um health healthcare has like eight times more data than the next largest enterprise industry on the on the in the country and over over 90% of that data um is unstructured. So these are all like sort of you know uh contracts and PDFs and handwritten notes and text that's like floating around the the the IT systems that exist. Um so it's it's obviously a massive uh opportunity and and problem at hand.

Yeah. Well uh congratulations on the progress. Give us the details about the round.

Yeah. Uh we raised a $38 million uh series

um uh fantastic uh led by the Pak 15 team which is a Sequoia India uh group along with General Catalyst um as well as uh YC and uh and a bunch of others. So uh uh yeah grateful for all the support and and uh interested in continuing to deploy pretty deeply.

When did you go through YC? Uh we actually started in summer 20. Um so yeah, I've been I've been sort of uh in the game grinding uh for the last uh I love it five and a half plus years.

Yeah, but I mean what you you caught the perfect inflection point. I'm sure you've done a ton of work to set up for the success and take advantage of the