Podcast

Dhaval:
Hey, Zach, tell us what you’ve been up to.

Zach:
So everybody, I’m Zach Hanson. Dhaval it’s great to see you again. You and I used to work together in a, past life in the AI field but What I’ve been up to lately is pushing for AI innovation within video, specifically at Brightcove, which is a great company, and working around how we actually build out and better experiences for our customers in the video space.

Dhaval:
Oh, wow. That’s like the most cutting edge space in AI right now. The video and AI generated videos and all of that as we speak in April of 2023. What specifically is Brightcove’s mission? And, yeah, if you could share a little bit about the, what is the product, what pain is it solving, and what are your customers, who are your customers? And then a little bit about where you are in the product space, like in the journey. Like, are you a startup? Are you, have you already found the product market fit? Are you enterprise? So yeah, any of that? So, a lot of questions, but just trying to understand the product story.

Zach:
So it is interesting, right? So we’re a very unique company. So Brightcove’s goal, to answer your first question, has become the most trusted company in streaming. So that’s our goal, right? That’s the big headline. That’s what we’re pushing and we kind of run the line of being a media company and an enterprise company. So from a competitive landscape perspective just to frame where we’re at, would be the Vimeos of the world, Kaltura and some other companies that are providing OTT services to brands around the world. And Brightcove has actually been in the game for over 15 years. So to answer another piece of your question is we’re a publicly traded company. We’ve been around for over 15 years and we’ve been providing these services for live streaming, for video on demand OTT for years, and it’s pretty amazing. So we’re one of the biggest little companies you might not have heard of. But with that comes a lot of responsibility around data because we actually ingest around one to two petabytes of video data every month. So we have an absolutely enormous catalog and data warehouse of videos, audio, all sorts of content that is being leveraged by our customers. Now, to answer one of the other questions rolled in there about our customers and some of the ones we can talk about, we help deliver video, great quality video with our encoders for the Olympics the year ago. Wow. Yeah, we’ve done that. We work with south by Southwest. So folks, if you’ve watched conference video from there, that’s Brightcove under the hood masterclass for instance. So we have a huge list of really amazing clients who are doing All sorts of different things with video. And that’s what we’re trying to enable. Now, the other piece of your question, where are we at in the product journey? Are we a startup? Are we a mature company? And I would say from the delivery of video, we’re very much a mature company. But when we start to think about machine learning and leveraging Models that are out there building our own models, we’re really much more in that startup phase where we’re trying to find the appropriate product market fit for the different types of models we might build or leverage the really immense amount of data we have to train models and do some really cool things.

Dhaval:
Wow. Thank you. Thank you so much for answering all those questions. I threw a barrage of questions at you. Wanna dive in a little bit on your last answer here on the topic of being new to AI ml. And what I wanna understand, Zach, is what is the customer trying to do when they want to use the AI ML capabilities for Brightcove? What is the thing that’s going on in their head when they are like trying to use your specifically AI ML features?

Zach:
So this is where it’s also like an interesting story because there’s a lot of stuff that we’re focused on from a machine learning perspective, kind of under the hood, things that our customers might not know is being powered by machine learning. So some of that has to do with encoding and how we get the video to the actual end user in a very efficient manner or in an efficient manner. As far as doing CDN optimization and making sure that the ultimate end user, which is our customer’s customer, whoever’s watching video. Has a great experience and that’s where the bulk of our effort’s been. But when you think about pain points, as we think about becoming more of a media company, when we think about enabling producers of content to be able to do some really cool things there’s really this kind of crawl, walk, run approach. One is when somebody uploads content to our catalog or their catalog through Brightcove, there are sorts of metadata that should be tagged in those videos. Oftentimes people are having to do that manually time stamping stuff or putting this as a certain piece of a sub catalog within their overall experience. So we’re trying to do some automation through their of automating tag management to suggest to our customers tags they might need in order to ease the burden of some of the metadata management, but then you go up the chain to content itself, the video, and we start to think about object recognition and video. We start to think about segmenting video. So you can easily cut and pull out specific elements of a video. For instance, if you were watching a soccer match or football match I grew up in the United States, so I’m a little bit more used to American football and I’ve become a bit more of a fan of the universal football in the years, in the past few years. But you might have an hour and a half long game and only have two goals or none. So the ability to be able to search through a video and find that really intense moment where somebody actually scores a goal, be able to rip that out really quickly and repurpose that content for marketing is very powerful. And there’s a lot of startups actually playing in that space and then you have the bigger players like ourselves, Brightcove. Then you also trying to play around with segmentation of video. So it really runs the whole gamut where we’ve been focused mostly on backend support, leveraging ml. All the way to that kind of front facing customer content production type of use case.

Dhaval:
Yeah, I, that’s amazing. You have, I can think of so many use cases. I was at a photo shoot video shoot event this weekend where I was hosting it, and we have like terabytes of video content that we created and now I want to create recap videos for that event. And I can imagine being able to feed something like that to your platform. Is that, am I getting it right? Would that be a potential use case? Is that how you It is parts out valuable clips.

Zach:
Exactly. And that is part of a potential use case that we’re exploring. But this is where it goes back to being in that kind of pseudo startup space. Like with all the data we have. With the great customers that we have, there’s a lot of opportunity there and we’re still in that feeling out phase of saying, what are those pain points to your other question. And like the use case you just gave, that might be something at the top of mind for a lot of our customers. And that’s where we’re just starting to put the feelers out and understand how we might be able to build some of these things out and make sure we have the right product market fit before rolling something out to our broader customer base.

Dhaval:
Yeah, that’s very interesting. Just like thinking for like content creators like myself. That event is an example of a use case. This podcast is an example of a use case, parsing out insights from this video, insights, and then publishing them. And then for courses that I create on product management and artificial intelligence, it’s the same thing. I can imagine being able to give you a whole course and create a trailer for that. So there are a million use cases that you can be going after. Are you thinking of like any big use cases right now that you are willing to share with us, that you may wanna pursue in near future?

Zach:
You know, none that I want to talk about specifically for Brightcove. But there are a lot of really interesting things in that segmentation space that just interest me and that might be wrapped into something we end up doing with Brightcove. It might not, but meta just came out with a paper on Segment. Anything. Have you heard of this model that they’ve built? Again, it’s in that segmentation space on video or, things like that. But with all the other models, stable diffusion. People are starting to piecemeal these different models together to come up with really cool use cases of really just actual content generation where you could take freeform speech to say what it is that you want to put together. Like for instance, it could be a thumbnail. Even a video thumbnail for the conversation we’re having now and with stable diffusion, with Segment, anything you might be able to pull an appropriate clip, you might be able to change it to meet and give a visual for this conversation that you know, you would normally have to go and pay an artist or get a graphic designer to pull together in the snap of a finger. So there’s a lot of Git repositories out there that’s like edit anything which pieces all these models together and they’re just a lot of really interesting work being done there in this space. That makes me excited and makes me excited to see how we might be able to leverage these different technologies just in video in general.

Dhaval:
Yeah. You are in a hot space, like what you just said is like so much happening there and being able to. Creators are living in such a good time right now. Like they can have great idea, great creative ideas, and the execution is streamlined to the max. Almost to the max, right? So as we go forward with this progression of technology in this space, Zach, Do you envision a future where creativity is a lot more valuable thing? Or do you see that being, co-piloted with ai. What is the value of creativity as we progress with this technology that is trying to commoditize it at the same time?

Zach:
Man, that is a philosophical question that is hard to answer now. I consider myself a creative, like you can see behind me, I like to write books. So, from that perspective, I like to flex the creative part of my own brain with coming up with stories for fiction books or writing non-fiction books. And as ChatGPT has come around, I of course have kind of played like giving certain prompts to see if it would give a similar storyline as to something that I’ve come up with. And of course, When you prompt it in the right way, it can get pretty close, but there’s always a little bit of a gap. I think there is a premium on human creativity that, I’m not as bullish on thinking that it’s going to disappear with the advent of a lot of these great technologies. I think there will always be that value for that human perspective. But like with GitHub co-pilot, with ChatGPT, I think what is happening and what will happen is the. Ramp up or the curve to get creative content out there will probably be reduced. Meaning if you can leverage it in an appropriate way to bust through rider’s block to Refine a story a little bit more quickly than you would if it was just pen and paper. Same with code. If you can leverage co-pilot in a way that allows you to get 80% there and then really buckle down to make sure you have the finer points really well fleshed out, I think it’s amazing. So I think it’s going to actually more unlock the ability for people to be creative. A lot more at least more during the day. So I think it’s actually a great tool. So I’m optimistic, I’m not a pessimist.

Dhaval:
That’s great, man. I think in the same way, I think as this tools advance, we will be seeing, a breakthrough in creativity for people who don’t think they’re creative. It’s gonna create that lane for people they’re not necessarily seeing themselves as creators or creative people. It will enable and empower them to be creative and we’ll see a whole breakthrough in creative potential of humanity as this unfolds. I’m looking forward to that future myself.

Zach:
And, on top of that, so similar to you, like I teach a course in data product management at Boise State University yesterday in my class, which is all senior level software engineers, computer science students. A lot of them are working right now and just doing school part-time. And one of my students talked about chatGPT, and he had a non-technical product manager on his team who ended up using copilot to just try to explore a little bit more and understand what his team was doing. And he said it was actually amazing, was co-pilot. He was able to get 80% there. Now the product manager lacked a few of the key core computer science fundamentals to get there. But it was just a little bit of a push from my student and his co-worker to get them all the way. And he was like, it was amazing. He’s like, now our product manager has a better understanding of what it is we’re doing. So I think it is also another opportunity for product managers and as product HQ to be able to. Empathize with their developers, more empathize with machine learning engineers because they can go out and explore on their own with a little bit more ease. And I think that’s an amazing potentiality for growth in the product space.

Dhaval:
Yeah, I love what you just shared. Thank you so much. Changing the gears a little bit, what are some of the product market fit lessons you are learning as Brightcove? Or any of your other advantages is starting to integrate AI into your product.

Zach:
Yeah. I have a lot of these just from years past. Right. Overall, A lot of companies and you and I have been in this game for a decade, right? So it’s now becoming I think easier to integrate ML into products, expose that to customers and have them appreciate it. But, my experience has been in the past five, 10 years is that oftentimes. Coming in as a machine learning product manager, you tend to push for things that are machine learning related, when oftentimes a simple set of rules might do the exact same thing and be more cost effective. So I think in past roles I’ve pushed for ML when really it might not be needed, or we might be a little bit ahead of the curve. So I think there’s even today something to be said about. Hitting the pause button as you’re thinking about implementing ML into any product you have. Like, is that the right solution? And oftentimes now it might be, but taking a step back and saying, can we do this more simply in a more cost effective way, that’s not gonna take up more compute power. It’s not going to need to have a whole model development and deployment pipeline tied to it. So, thinking about it from that perspective and making sure the market is actually looking for an ML solution is key. Because I’ve been hung up on that several times in the past where we overbuilt for something that was actually quite simple.

Dhaval:
Yeah. And like you said, now it may make sense to. Do ML or AI because of the ease of doing it. Right even then though, like, I like what you just said, which is still take a step back and see if you really need it. Are you just getting sucked into the hype or does your customer really benefit from you doing it? And that’s an important distinction, like that’s a really important distinction you are making there. Thank you for that. What are some of the, specific challenges that you are running into in terms of working at this large organization that is trying to find this new foothold with this highly innovative use cases? You may reframe them as opportunities rather than challenges. But yeah, however you can, share that. I would love to hear that.

Zach:
Yeah. The answer is data, data, data, data. And you know, it is both opportunity and it’s both a little bit of a road blocker. companies that have been born into the cloud era that have been born with data first principles you can see are really getting ahead right in the ML space because they have. Well kept data with great metadata tied to it. They have strong taxonomies. It’s well annotated so you can build models relatively easy or build features to put into models and it’s great. One of the things that I’ve struggled with across the board and a lot of the companies and I’ve been a big financial institutions even with Brightcove, that’s a 15 year old company now. Is making sure that we’re treating data as a product. That’s a big thing for me. It takes part of the data mesh principles, which is if you are building a product and deploying it within a company, it could be Brightcove, it could be IBM it could be anywhere. Oftentimes, that product that you’re gonna be developing is also gonna have data as an output. And that cannot be an afterthought because what happens is things start to metastasize and you get bigger and you move on to the next product, and then that produces data. And if you don’t have technical product managers and data engineers really evaluating the outputs of those data. And making sure that it has a strong taxonomy, making sure that the metadata tied to it is clear, concise, usable, traceable. Then you’re gonna be at a disadvantage down the road when you actually want to implement machine learning practices and machine learning models. Because you’re gonna spend a lot of your time going back and trying to clean that data or introducing a third party vendor to come in and either clean it, annotate it, and get it in a position where you and your machine learning engineers can take advantage of it. So for me, it’s all about data, and that is an opportunity for everybody, is to sit back, look at the data that your product’s already produced, and what you have available to build off of, and making sure that it is as clean as you can possibly have it. Implementing governance across the board to make sure your teams are treating that data well. And that way you’ve heard the, the trope about now people have data lakes and then often turning into data swamps because of that lack of governance. You wanna avoid that. And that’s the biggest opportunity that we have. And the biggest opportunity, I think, really any company across the board could have if you want to take advantage of machine learning practices as you grow.

Dhaval:
Yeah, data is still as relevant as it was 10 years ago and it’s never going to be less relevant no matter how advanced AI gets. So one follow up question is slightly, slightly different question rather, is, what are some of the opportunities for. Brand new people like trying to get into this space who don’t have prior AI ML background. I know you have been working in the space for 10 years. You’ve been doing this for, before. It was cool before you and anyone knew that there was something called AI product management. You were an AI product manager. Right? But for people who are just like oh this is not going away. I’m stuck with this field forever. Now I better learn it. It’s like it’s the new software. Right? So what are your. Recommendations for folks like that who want to learn about getting into AI product management or AI product creation space?

Zach:
Yeah, first and foremost, I got into AI product management by happenstance. You know, for me it started out as being an entrepreneur. So I still think the greatest base for getting into product management, whether that’s a data product manager, an AI product manager, is go build something. Get down core product fundamentals, which is for me, design thinking, understanding how to go from a broad problem space, refining that in the personas that are affected by it, understanding what they’re doing today, what you might be able to change to make their lives better, and then executing on a product. So that’s was my journey, was building a product that ultimately failed. Before I went into industry and then found my way into the machine learning space. But number one, build something. It doesn’t have to be a unicorn. It can be something very small that either sell or give to friends, family, whatever. Build something on top of that. If you are keen on making sure that you are gonna be doing ai product management is get comfortable with data. Again, we just talked about data being the core of everything. Yamak Dajani has a great book out around data mesh that covers a lot of the principles that roll over into machine learning and ai. It’s making sure you have federated governance around your data. It’s treating data as a product, so applying product fundamentals to the data you might use in machine learning. It has all the other elements that you might need to be a good AI product manager with core data fundamentals. So I think those two things are what I would suggest anybody getting into it. So build something and really understand data.

Dhaval:
Thank you. Thank you so much Zach. It’s pleasure to have you on the show. Looking forward to have you back on the show when we have more things to share and when you publish your next book.

Zach:
We look forward to it. Dhaval