Accelerating Safety-Critical Software Development: Erez Kaminski, Founder of Ketryx
Ketryx is an AI-driven developer tool that helps safety-critical software teams (medical devices, automotive, aerospace, etc.) develop safer software faster.
Ketryx is an AI-driven developer tool that helps safety-critical software teams (medical devices, automotive, aerospace, etc.) develop safer software faster. Ubiquity first invested in Ketryx’s 2021 seed round. In this interview, CEO and founder Erez Kaminski talks about the complexities surrounding regulated software development and why his mission to improve medical device software development is so personal.
Can you sum up what Ketryx does in one sentence?
Ketryx (www.ketryx.com) empowers teams to build safe and reliable software for highly-regulated industries such as medical technology, pharmaceutical, automotive, and aerospace.
Tell us more about Ketryx’s product.
Ketryx is the first and only connected lifecycle management software for medical devices and other regulated industries that unifies quality and R&D efforts to deliver safer, more innovative software up to 2-3x faster. Compliant by design, Ketryx builds guardrails to eliminate the manual, error-prone cutting and pasting from traditional disconnected medical systems. This reduces the risk of potential defects while delivering enhanced patient outcomes. It's hard to build software that is as safe and reliable as the FDA would expect, and we make it much easier to do.
What is the story behind the founding of Ketryx?
My mother is a physician, but she’s also a patient who has an implantable medical device. Her benefit from the technology is enormous. It has changed her life completely, and she is grateful every day for being born into this era. Nevertheless I have always felt it could be better. In many cases, medical device software doesn’t function as well as you would hope. It is missing features one would expect, especially given the risks patients take in surgery. That’s one side of the story: This issue is very personal for me. The ability to make safety-critical software easier to fix and maintain is a pressing need for society. When you talk to big companies about why they don’t fix their software, the answer is surprising: it’s because it’s really, really hard to update and takes such a long time to produce all of the evidence (i.e. documents) required by different regulators in different countries.
The other reason for founding Ketryx was out of frustration from my own work experiences: I started to work as a developer and AI expert. I used to work for a company that builds software development tools called Wolfram Research - they build tools like Wolfram|Alpha and Mathematica. Then, I was hired to be the Head of AI and Machine Learning for the medical device division of Amgen, the largest biotechnology company in the world that manufactures some of the highest-risk products ever made: injectable biologics. Its famous drug is given mainly for cancer patients after chemotherapy.
When I saw how people build FDA-regulated software, I suddenly understood why it is so hard to update medical devices. The way the medical device industry works is so burdensome and the tools they use were built 30+ years ago. At the same time, I understood why they had to jump through all these hoops. While there's a good reason the FDA has these expectations, the last few years of web-based software and machine learning software development is no longer aligned with this approach for safe and reliable software and medical devices. Ketryx is unifying those worlds: the ability to build safe and reliable medical software but with all the advantages of modern tools, including speed and greater accuracy.
How did you notice the need for an FDA-grade medical software development platform?
I joined a company that was engaged in building complex products. It’s hard to imagine what it takes to build a biological asset - the most modern drug (biotech) - and inject it into a person and to make sure that the device that's injecting it performs exactly as expected. Seeing how this is actually done was a bit shocking, and it isn’t anything to do with them - the industry best practices are not aligned with how it’s done in other industries.
I came from a company that builds development tools, so I see the world through the lens of tooling; the reason you’re hammering it wrong is because you have the wrong kind of hammer, or you actually need a drill. Not only is this true for medical devices, it’s also true for most highly-regulated industries. It's hard to innovate in automotive: 300,000 Tesla cars were recalled a few weeks ago, and the 737 Max Boeing incident was fundamentally a software failure. There are a lot of issues in building highly-regulated, high-risk software that deter people from working on them. And the tools look like that - as if someone's neglected this space - but we're all depending on it, so it's kind of sad.
That’s the founding story: seeing this problem over and over and saying, “If we spend some time and bring the right team together, we could really solve that for a lot of people.”
What attempts did you and your team try before getting to the current solution?
Well, we've been working on this one for quite a while, and the problem we’re trying to solve has been very specific: how to build regulated products—medical products—faster and cheaper. We were lucky to have industry experts and great partners at large medical device companies that helped guide our work. I think we went deeper and deeper into regulation and the rigor in which products are evaluated and eventually got to the highest rigor (outside of nuclear reactors), which is medical devices and pharmaceuticals. Because there’s so much value there - and because what the FDA really regulates is the value - the more value you promise the American public, the more regulated that thing is likely going to be. It’s not always the case, but it's usually the case.
Early on, there was another related avenue that we thought about. It was, again, helping people create safe products but through creating artificial intelligence systems that would process complaints from the public and help medical manufacturers understand how the complaints about their product tied to certain risks and issues in their product. Funnily enough, we ended up incorporating that into the current system we’re building, but that was the previous idea. It's all very much about how do you make medical products better?
How do you know customers love your product?
Within large and complex companies it takes a LOT of effort to buy things, so even though our product is highly scrutinized, our customers are championing very hard to purchase and use Ketryx because it's just of such great value to them. The fact they push through such friction to deploy Ketryx tells us we are onto something.
When did you first get into this particular area of safety-critical product development?
I’m an Israeli Defense Forces military veteran, and even in the military I spent time working on safety. So it’s a small world, and it all comes back eventually.
I moved to the U.S. where I studied physics and math in college. I was working as a physicist for a short period of time in a lab up at Princeton, and then I wanted to go work in industry. I went to work for this amazing company called Wolfram Research that is kind of the mecca of mathematical software tool building; when people say, “I am a professional mathematician, and I use special software tools to do math,” they mean this tool. So it was really exciting to be in a place that thinks a lot about high-quality, extremely complex software. Not much software is much more complicated than what they do there at Wolfram Research.
When I was first introduced to safety-critical software development is when I was offered to come work at Amgen from Wolfram Research and to come lead AI and machine learning for their medical device group. I was introduced to how they develop and just the way the system looked for this type of product development. I was very dissatisfied; it just seems like it's hard enough to do medicine and to develop medical products - why add that the tools are really bad to use?
Where did you start as a kid? What drove you?
I think being the son of my mom is a reason why. My mom isn’t only a patient for this particular type of device - the cochlear implant (a neural prosthesis replacing certain sensors connected to the human brain that allows you to hear). She’s also a doctor who performs that type of surgery. Over time, I realized that one of the key inhibitors to broader adoption of these and other devices is the fact that the software tools are totally broken…that's a big motivation to solve that.
It's not an easy problem to solve. It’s a big hairy problem, and there's a reason it is like that. We have a lot of motivation to solve it because we kind of lived with this, and it's not just me. We have a whole team of folks who have lived with this, from people who built huge systems with me to folks who wrote the American regulations. We have a former FDA official who joined us out of retirement who wrote most of the American regulation that pertains to medical device software, and he joined us because he's thinking we’re at a breaking point where it's almost impossible to build really, really big and complicated safety-critical systems.
We think of nerds as people who are obsessed with something. What are you nerdy about or obsessed with?
This field is really my life, and I think it’s such a fascinating problem. Why is it so hard to prove that a computer system or product does what you said it does over a massive-scale supply system of hundreds of people designing it, thousands of people involved in manufacturing, and thousands of people involved in shipping? It’s a really interesting puzzle that I'm excited to work with our team to solve.
I also like physics. I used to do a lot of physics, and I still read about plasma and fusion physics. I think the new fusion reactor news is pretty exciting. And I'm pretty geeky about football, too. I originally moved to the U.S. to play football in college in the Deep South in Birmingham, Alabama. I still watch football and geek out about that - lots of statistics to analyze.
What have you gotten out of your relationship with Ubiquity Ventures?
A partner who understands people trying to do big things and who is helping them do big, complicated things with the proper type of financial and strategic support.
What keeps you going during tough times?
That we're doing something that's important and is important to a lot of people. And having a team that's uniquely equipped to solve it and knowing that we're going to solve it together.
What would you tell your past self if you could give them advice?
You can never move fast enough. As fast as you think you’re going, you should always go faster and you would retrospectively want to always go faster.
My words of wisdom: make sure you understand what you're getting into, and building things you care about is really important. There are a lot of tough days, and building something that you care about personally and that you know is going to help people helps. And it’s not just about being cool and helpful, it also has to have a market. I think if you can do that and you can figure it out, it’s a lot of fun. So moving faster and making sure you work on something you care about is my advice.
What’s your advice to budding technical founders who haven’t yet taken the leap to launch their new company?
Make sure you're building a great team around you; you need a village to do this. You always need to just try it out and then see. If you do have a deep passion and you feel like you have an insight over something, especially if it's very complicated and tough and difficult to do, that's all the more reason to just go do it.
Ubiquity Ventures — led by Sunil Nagaraj — is a seed-stage venture capital firm focused on startups solving real-world physical problems with "software beyond the screen", often using smart hardware or machine learning.
Are you pushing software beyond the screen? Using smart hardware or machine learning to do something in or understand the real world? Let’s talk!