Kate Migliazzo

I run product for Native Frame, which is a live-streaming infrastructure company. Our customers are typically small to medium-sized businesses who see video as a core feature of their product. But video being so specialized, they prefer not to build it themselves. So that's where we come in.

Yeah, anything that they want to build native into their product. A good example of a customer that we have a long-standing relationship with is a customer called Replayer. They're based out of LA. They build autonomous cameras for youth sports. So if you're a parent that's live-streaming, or a coach that's live-streaming or recording that game, we're sort of the streaming engine under the hood.

Yeah, we've met with them.

Yeah, a much smaller FloSports would be a type of customer that we would work with.

Yeah. From a product standpoint, we design for three user types. The first one being the broadcaster — the person who's sitting behind the camera. Could be a parent, could be an event organizer. They're the ones who's going to hit go-live. The second is the viewer, the fans, the audience, those that are consuming that content. And the third is going to be the developer who's integrating our APIs into their platform. So Claude is sort of our go-to tool for now.

Where I've found it extremely helpful is to be able to continuously build on these profiles. Whenever we're trying to solve a specific user problem, it can help us flag for any inconsistencies, or where we're leaning too far into one user problem, and what are the trade-offs that we're going to be having to work around.

Yes, absolutely. Because the engineering has become so much more efficient, product and design have to keep up with the same pace. Even doing user research and creating artifacts around users, it can go from a process that used to take weeks to days, hours.

That is a great question. I think because we have so many cycles that we can iterate on — at what point is something in a place where we can move forward with it? Because you are layering different tools, different models, but also the human element. I think it's the human part, really — being able to look at whatever output you're getting and evaluate it, and then think about it from feasibility and the actual implementation of something.

You put it in front of the user. Those prototypes that, like I said, used to take weeks or days to create, we can create something that's clickable. We can show it to a user and we get their immediate feedback, which is part of what we live for — being able to see the instantaneous feedback, whether it's positive or not so positive.

Yeah, it does create a bit of — you know, where you're working a lot longer than you should be, because there's just this volume of ideas. There's always something to explore, something to pick at. I think that's become a problem for me, just a work-life balance.

To be honest, I don't think it's changed that much. Of course, it will help expedite looking through portfolios and resumes. But when you're meeting a person and you're talking with them in a live format, you're still looking for the same attributes of someone who is a self-starter, can think about an objective, and can work autonomously toward that objective. Someone who's just good to work with as a person.

I think it's a little bit — just because the velocity changes doesn't mean that what we want to accomplish decreases. In fact, it increases. Your aspirations get loftier. So I don't know whether that's a good or bad thing. But hiring, we're continuing to hire. I think we're all trying to become more fluent with these tools — someone who understands how to use them, who has an attitude of being able to experiment and being able to combine, and again, being able to bring a human element to the process.

We have about 40 developers.

Yes, the velocity has increased tremendously. I'll let them speak in terms of the quality — I'm not going to necessarily judge that. But our firm is engineering-driven, so there's a lot of discussion around setting up different processes in terms of reviewing code and creating code, and migrating from this identity of a developer to a developer who knows how to become more efficient and smarter at their work. It's really interesting — there's a lot of enthusiasm for AI, which, you know, you never know what to expect when it comes to something that could potentially take your job. But I think it's fully embraced on the teams that we have, and people are finding new and creative ways to use it.

Yeah, we have a combination of tools. We are in the process of building our own project management tool where we're moving all of our tickets into components that really are more related to a business objective. Whereas before we had lots of small technical tasks that were very difficult to map to overall objectives. So we're able to customize our own process. I think we do things a little bit differently because we are smaller, and we have different types of stakeholders, so we have different reporting needs.

Physically we have a hub in Austin, but a lot of our core team members are in different time zones. So a lot of our communication — despite the fact that there's a concentration of us in Austin — we still conduct either async, or over video calls, or Slack.

No, they're all internal.

Certainly our customers — we have customers in Asia, Australia. We have developers in Tel Aviv. So it is very much a remote, virtual type of work environment.

I enjoy not having to do some of the documentation that is required with product — being able to create requirement documents. A lot of the standard product work that I don't like to do is automated for me now. So I'm really excited about that. A lot of people get into product because we're interested in the end user and the user experience and solving that problem for them. When you spend a lot of your day writing out detailed requirements, it sucks the energy out of that process.

That's a great question. I do find myself defaulting to AI when I'm feeling lazy or when I'm in a hurry. There is a cost to that cognitively, and to building my own knowledge base of what's actually happening. So yeah, I do think about it.

I try to devote my attention to the problem at hand. If I'm close to the problem, if I can devote more of my attention and energy to where I think it matters more, then I feel like I can stay relevant to what the team needs from me.

When you mentioned hiring — the one thing that does keep me up at night is, we're not hiring as many junior people. This can be developers, this can be product people, designers. We tend to hire for senior-level employees. I think about the sustainability of our operation and being able to ensure that a lot of the expertise and the knowledge that our senior-level people have — being able to train the next cycle of people. I wonder when we're going to hit that moment of, everybody's retiring or moving on, and we've got all of our skills and expertise going with it.

...just going to retire in our big house.

Yeah, it's something that we do think about a lot, because right now, for our users, video is extremely labor-intensive. Our core, our grassroots community — passion activities where your end user is going to be someone who's using a consumer-grade device. They're not going to have the professional equipment or the lighting. They're just going to want to pick up their phone and hit go. So there's that piece of it, but there's also: they don't want to have to flip through reels and cut clips. They want that to be done automatically.

So we think a lot about how AI models fit into our pipeline — the video pipeline — and that's one thing that we are designing for. It's to be invisible, really, to the user. The experience is, it's just done for you. Our pipeline is designed so that AI as a model is a native component to our pipeline.

Today, the way that a lot of AI models work with video is that you capture the video, and then the AI model lives in a different place. That video gets shipped to the AI model, it does its processing, and it comes back. That can take — sometimes it could take days, sometimes minutes. What we've thought about is being able to treat those AI models as a first-class citizen — so they're sitting right next to the video and everything is happening in real time. It may not matter for a lot of cases — if you're creating a video reel for highlights for athletes, recruitment, the time doesn't necessarily matter. But if you're a parent on the sidelines and you're watching your kid make this unbelievable move, you want to be able to instantaneously clip that and share it with your family. So we're building a pipeline that can enable that.