Medra raises $52M Series A to build autonomous AI scientists that run lab experiments at scale

what's going on

the TV van alum. Welcome to the show.

Hey guys.

Hey.

Uh, excited to be here.

Thank you so much for having you and your robots

punctuality. Oh yes. What is behind you? Wow, there's a robot that's actively working. Explain, introduce yourself, please.

Absolutely. Uh, so I'm Michelle. I'm the founder and CEO of Medra. And a little bit about me.

Yeah.

Uh, I come I studied chemical engineering in undergrad. Was your typical chemistry life science nerd? And then I did an internship at SpaceX.

Cool.

And was just really excited about like what if we can build in the physical world.

Yeah.

Right. Like I wanted to build in the physical world. I want to build real things with real impact.

Yeah.

Ended up doing my PhD at Stanford at the Stanford AI lab in robotics, building robotics foundation models. I worked with Janette Vogue and also with Fe Lee. Uh, shout out world labs. say

and uh I ended up when I finished my PhD decided I wanted to combine life sciences, robotics, AI and I started Medra and we are building physical AI scientists which we

think that is necessary to eradicate disease.

How narrow do you want to go to start? I mean it feels like there's pipe heading, there's centrifuging, there's different stuff going on behind you. Uh but like medicine, bio, these are massive terms can be animal studies, mice models, you could have monkeys in there. There's a million things that you could do. Uh but I feel like you you probably want to pick a beach head, but you tell me what the strategy is.

Definitely. Look, like one day we will have metro robots doing animal studies. Like mark my words, right? Uh but you're right. We have to start somewhere and we're starting with early discovery and development. We have physical AI robots at Medra that can do experiments at scale. We can work with

instruments that humans already can use. And most importantly, we truly have intelligent robotics. This is not just lab automation where you program things and they do it exactly like uh you tell it to do. This is actually uh physical AI autonomy that is intelligent that's constantly sensing making corrections and more importantly we also have an AI scientist that can actually reason about the science itself.

So what is an example in the lab where you actually do want some probabilistic reasoning or some stoastic result as opposed to something deterministic? because if I'm if I'm vibe coding a website, like I don't want it to guess what an HTML tag is. I want it to just use a div every time. It does a great job at that. Uh but so I imagine there's some things where you know the pipet always needs to go in the same place. So it's okay to stand on the shoulders of giants and puppeteer that where does the variability come in?

Definitely. And I I think like if you think about the best scientists, right,

the best scientists are the ones who are reading all the papers. They have all the scientific knowledge, but they're also the ones going into the lab and running the experiments. They can like sense what's happening. They can smell it. They can like visualize what's going on and they can make changes as they see things um start happening inside the experiments.

That's what we're trying to capture, right? the ability to be really flexible to actually reason about the science as it it's happening and also taking into account all of the knowledge that's come before us all the scientific papers all the different results all the past experiments you run

that's actually what enables good science

so tell me about the distribution business model I could imagine a world where you're basically doing drug discovery going through the FDA process at the same time. Uh you could sort of sell a lab in a box to a pharmaceutical company. There's a lot of different ways I could see this taking shape. Where do you think this goes?

Yeah, we are building the infrastructure layer. We want to be the TSMC for drug discovery. So we are partnering closely with pharma companies, biotechs such as Janentech. uh where we are they can either work with us by using our system our physical AI scientists in their own lab or we're actually about to open our own lab our own fully autonomous lab one of the largest autonomous labs in the United States uh in 2026.

Talk to us about the fundraising. I think we missed you on the day you announced your series A, but I still want to ring the gong. What happened? Who's in How much did you raise? Yeah, we raised uh $52 million for series A.

Yeah. SC, thank you.

Amazing. Who who'd you raise it from?

Yeah. Uh human capital uh led uh they came in and preede and seed and they tripled down on us for series A. Really preempted the race. Uh we also have Lux who is also a repeat investor. Um also Melo Ventures, Catalio. uh great investors uh joining in for a very ambitious mission and very ambitious journey of eradicating disease

and is 52 million series A that's feels like a lot of money congratulations but is is I I could imagine spending it on a training run for a foundation model or buying a bunch of robots like that stuff behind you doesn't look too cheap where do you see the money going what does it unlock

well actually the hardware that we use at Medra is all off the shelf um robots right now especially their hardware is fairly commoditized and we use this off-the-shelf hardware so we can build AI on top of it so we can actually reason about the science and actually uh be able to use our uh what we have trained ourselves which is the vision language lab action model to be able to autonomously run experiments and a lot of what we have raised our series A for is actually to open our own lab

right in San Francisco uh to be able to scale up data generation because ultimately what want to do is to be a data foundry. Sure.

For life sciences to be like a merur search but for biological and life science and chemical uh chemistry data so that our partners can train foundation models in biology.

Yeah. Because there's probably not a lot of really clean data out there on GitHub or out on the open internet and so you have to sort of generate it yourself. Is that generally correct?

That's right. That's right. I mean, if you think about in biology, like the largest biology foundation models are still about three orders of magnitude trained on three orders of magnitude less data than like, you know, even like 01, which is

Yeah. Yeah. Yeah. I think Google launched one that was showed really impressive results and the scaling laws were there, but it was much smaller than what you see elsewhere. So, yeah, very interesting. Very exciting. Jordy, anything else?

No, this is uh super exciting.

Congratulations.

In the future,

and I'm sure we'll have you back on the show soon. Have a great rest of your day.

Great to meet you.

We'll talk to you soon.

Let me tell you about Shopify. Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in store, on mobile, on social, on marketplaces, and now with AI agents and

uh we got a little bit of time. I think

we do. That's great.

Our next one.

Well, then I'll tell people about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange.

And let's pull up this post from Gucci.