2.28-29.24-Ethics-Ketryx

EREZ KAMINSKI: Nice to see you all, and thank you for the Startup Exchange group, Catarina and Irina for having us here. Last time I was here, I started an engagement with a Fortune 500 company. So I hope today we'll see another resurgence of that. My name is Erez. I'm an MBA EECS master's student from MIT. Before that, I worked at Amgen in the AI numerical simulation space. Happy to see some former colleagues here. So that's nice.

And today, I want to talk to you about risk-based AI. So for folks here who work in regulated industries where their products can cause harm to individuals or society, there's a lot of regulations involved. I'm sure that folks here from the pharmaceutical, medical device, automotive, defense, and other spaces are quite aware of this problem, where when you make a highly regulated product, whether that's software or AI, machine learning just makes it more complicated to do, you need to do a lot of product-level compliance requirements.

This is something you need to do for every release of the product, right? This is kind of the core issue that causes highly regulated products to be released slower than nonregulated products. Like, most tech companies release software every 60 minutes. Most Fortune 500 automotive companies do not release products every 60 minutes. And probably the factor is like 100 or 1,000 difference.

The issue that comes up is basically it's just really hard to do these type of things. And the amount of effort involved is growing every year. If we look at that graph that other gentleman had earlier of going into spreadsheets, this is where we take the best and brightest in biology, mechanical engineering, electrical engineering, software design, and put them in a room with spreadsheets and software requirements, specifications, and other design control documents and risk management documents and make them do a lot of manual work.

We've automated all that manual work through a series of lifecycle management tools that connect to existing, commonly used software development tools like Jira, GitHub, and AWS. And we help companies all around the world to produce highly regulated software, including regulated AI, faster and cheaper.

So today teams need to decide, do I do modern DevOps or modern machine-learning ops? Or am I compliant? Because they can't find a way to do both. And as we all know working in highly regulated industries, we need to be very compliant from a safety perspective. Compliance is not about the law. It's about how do we not harm patients or other folks. That causes a lot of challenges.

And we don't think that teams need to make that decision. We think the teams today can use automation, a lot of machine learning, generative AI features that we've embedded into our tool to make it possible to release software and highly validated products at the same timeline that it takes to release regular products-- maybe a little bit more because you need to do some risk management and author some things but not to the scale that we're talking about today.

So Ketryx is how teams develop today risk-based AI, mostly for high-risk scenarios. But any AI or any software systems that require a combination of risk management, validation of the product, as well as a management under a quality management system, we found a way to automate both the rules of the quality management system and the generation of evidence from that quality management system, again, in the tools that every developer in the world already knows. They almost don't feel like they're working on highly regulated products.

There's a lot of risk-based AI out there. And these are just some of the applications most of them were already in. So AI-driven medical devices were quite broadly deployed there in infusion pumps, robotics, and things of that nature. We're looking to get into other industries like financial decision-making, autonomous vehicles, any other critical infrastructure that requires, again, QMS, risk management, and software validation.

You could find a huge benefit in our tools, reducing the time you spend making documentation by 50% or more. And not just that, making it possible to hire a developer from any place in the world and having it be productive inside a highly regulated, high-quality validated environment within weeks and not months.

I'm going to share a few of our private companies. We don't really disclose a lot of our public customers for obvious reasons. What we do with them is kind of a pretty big trade secret. If you can move 30 to 50% faster than any competitor, it's a big deal. But I could share that we're already deployed to over 1 million patients. And actually, we heard a few days ago there might be 5 million patients. We just don't know because we're not meddling in their affairs so much. But these are some of the top brands here in Cambridge that already work with us that folks know.

Very shortly, what's risk of AI? It just means that you have a use case you need to create a bunch of requirements for it, specification, code it, put it in a device or, if it's totally virtual, in software. And then you manage risks and post-market surveillance and efforts around that. We've kind of automated that complete process in a very, very unique system. And I'd say the challenges of this is the challenges that everybody faces that we solve. It's very hard to find people who know AI. You don't want them to spend their time doing documentation. It's very hard to find subject matter experts in your field of practice, medicine, biopharmaceutical design, financial services, and algorithms.

And then finding someone who understand the regulatory framework they need to conform to while using the type of tools they want to use in order to release fast in a CICD fashion is just seemingly impossible. We'd love to talk to you all. We have a booth right out there, first door on your left. We're the first booth on the right. You can always connect with me at erez@ketryx.com. And we have an office right around the corner, so we're happy to host people. And we're looking for leaders who want to change how they develop software, reduce their costs, and move much, much faster at the largest deployment on Earth. That's what we're trying to do. So thank you.