This episode is brought to you by Reka, a developer of industry-leading, multimodal, AI models that enable individuals and organizations to deploy generative AI applications.
Agility requires brands to utilize technology like AI in new and innovative ways to better understand information and interact with their customers, giving them what they want, when and how they want it.
To help me discuss this topic and some of the recent announcements from Reka, I’d like to welcome Meenal Nalwaya, Head of Product at Reka.
Meenal Nalwaya on LinkedIn: https://www.linkedin.com/in/meenalnalwaya/
Resources
Reka: https://www.reka.ai
This episode is brought to you by Landrum Talent Solutions, a national recruiting firm specializing in marketing and HR positions. https://www.landrumtalentsolutions.com
Don’t Miss MAICON 2025, October 14-16 in Cleveland – the event bringing together the brights minds and leading voices in AI. Use Code AGILE150 for $150 off registration. Go here to register: https://bit.ly/agile150
Connect with Greg on LinkedIn: https://www.linkedin.com/in/gregkihlstrom
Don’t miss a thing: get the latest episodes, sign up for our newsletter and more: https://www.theagilebrand.show
Check out The Agile Brand Guide website with articles, insights, and Martechipedia, the wiki for marketing technology: https://www.agilebrandguide.com
The Agile Brand is produced by Missing Link—a Latina-owned strategy-driven, creatively fueled production co-op. From ideation to creation, they craft human connections through intelligent, engaging and informative content. https://www.missinglink.company
Transcript
Greg Kihlstrom (00:00)
This episode is brought to you by Reka, a developer of industry leading multimodal AI models that enable individuals and organizations to deploy generative AI applications. Agility requires brands to utilize technology like AI in new and innovative ways to better understand information and interact with their customers, giving them what they want, when, where, and how they want it. To help me discuss this topic and some of the recent announcements from Reka, I’d like to welcome Meenal Nawaya, Head of Product at Reka. Meenal, welcome to the show.
Meenal Nalwaya (00:31)
Greg, great to be here.
Greg Kihlstrom (00:33)
Yeah, looking forward to talking about this with you. Before we dive in, why don’t you tell us a little bit about your background and what brought you to Reka?
Meenal Nalwaya (00:41)
Absolutely. Well, I’ve spent nearly the past decade of my career across Amazon and Meta focusing on converting really advanced technologies into products that are used by millions and billions of people across the world. And at Meta in my last stint, I worked on multimodal reasoning models for Llama and in Meta AI. And at that point, it kind of became clear to me that AI isn’t just about bigger models.
It’s really about building solutions that are solving real world problems. I also clearly saw a gap where while most of the world’s data is visual and audio, and which is how humans see the world as well, most of the AI that was being built was around text. And that’s where I saw an opportunity and I got, and I jumped ship and joined Reka.
Greg Kihlstrom (01:30)
Yeah, yeah, yeah, and definitely agree. I’m I’m I’m fascinated with this this topic. So looking forward to to learning a little bit more. And maybe for those that are not familiar with Reka I know we’ve had we’ve had a few episodes here with with with people from Reka. But for those that are less familiar, can you talk a little bit about what it is and what makes it different?
Meenal Nalwaya (01:52)
Absolutely. So at Reka we’re really building the enterprise standard for multimodal agents. And we are a full stack multimodal model as well as product company. And our differentiation here really comes from two key pillars. First, we’re multimodal first. So we build agents that can see, listen, and reason across the enterprise data that most of the AI has actually ignored till now.
So if you look at most of the AI adoption that’s happening around you, it’s really focused around using text data or using or coding has been highly transformational as well. But images and videos have not gotten that sort of love. And our focus is really around multimodal from that perspective. The second pillar is we are a full stack company. So we are not a wrapper company. We don’t use lean on OpenAI or Gemini’s models, but we build our own multimodal foundational models and then package them into solutions that can be used by our customers. This gives us a lot of control over the kind of quality, customization, and the speed with which we can move because we’re not dependent on anybody else’s roadmap. And that’s kind of fundamental to our ⁓ company and our product strategy.
Greg Kihlstrom (03:09)
Yeah, yeah. And so you’ve had some exciting launches recently. I’d love to hear some highlights of those. Why don’t you talk through some of those? Yes.
Meenal Nalwaya (03:18)
Yes, very excited to share those. We have launched two key platform capabilities. First is ⁓ Rekka Vision, which is a platform capability that lets enterprises have a human-like understanding of their images and videos. What this does is it allows you to ask questions and search over your videos and images directly. For example, you can just say, show me every slam dunk from last night and it went instantly generate a highlight for you. I believe, Greg, you have also tried some of our apps on that. Or as a security person, can say, find me all security events this week and generate a report for me. Something that security officers don’t want to do because it’s very time consuming for them. And it will generate all of the, it will find all of the relevant clips and it will generate a structured log for you for any of your compliance purposes so that you can really focus on what you do best.
The second platform capability we launched is RecurResearch, which adds a reasoning layer on top of this. So now we can go beyond search and conduct deep research not on all of your private enterprise data, as well as across web. So really build that agentic component. It provides structured and cited answers so that your team can really just go and trace back any of the output and really have this transparency and trust element built into it. So for example, you can ask what teams are trending this season versus last year and will give you a boardroom-ready report with all the citations that you can go back and check on. So together, if you combine vision and research as horizontal platform capabilities, it kind of brings together the perception side of the world and then adds the reasoning layer on top of that, which are sort of the building blocks for multimodal agents in any industry.
Greg Kihlstrom (05:07)
I wonder if you could give us maybe a real world example of this stuff in action.
Meenal Nalwaya (05:13)
Yes, given that multimodal is really our key pillar of product strategy, we really focus on industries where unlocking this multimodal data is critical. One of these industries is security. In security industry, can imagine there is a that, know, we can imagine multiple hundreds of cameras and a large amount of security footage that these officers are overwhelmed with analyzing and looking through.
So for example, we worked with the Ohio Police Department where we implemented Reka solutions across more than 100 intersections that they have. And now with these Reka powered agents, they are able to achieve 65 % faster case resolution and 42 % reduction in crime. That is crazy amount of improvement in safety and speed of resolution, which I don’t think any other tech can achieve at this pace.
And this wasn’t just about cool tech for us, it was also about solving a real workflow bottleneck and delivering safety for all of the citizens. So as police officers, the response we hear from them is this allows them to really save time. They have achieved 40 % savings in officer’s time so that these officers can really do their job and not spend time finding the right clip and creating what report and doing compliance.
Another example I would love to give is media as an industry. Media, as you can imagine, the content studios and broadcasters have huge archives of videos that are sitting around, many, many years of content that’s there. But it’s very hard for them to search and find any relevant footage. And then if you can’t find it, you can’t repurpose it, you can’t repackage it, and you can’t monetize it.
So now with Rekha’s platform, a media company, for example Paramount, can transform its business in multiple different ways. For example, one where it comes to licensing. They might have decades of content sitting in archives. And now with our human-like video understanding, they can make every scene searchable at the scene, theme, actor, even emotion level. And now if you can search and discover footage faster, finally every time Brad Pitt was shown in a video, you can license the content a lot more efficiently. And not only that, I think this can also transform how content is discovered across the platforms. So for example, if Paramount licenses this content to partners, let’s say it gives it to Netflix, then Reka can drastically just change how the discovery of this content works. Now that you have such rich metadata associated with this content, it will be much easier even for the consumers on these platforms to find content. And then what better discovery means more engagement with the content. And you just get more bang for your buck and all of your licensing deals. For example, imagine on Netflix, instead of just searching, find me a romantic comedy, which I think is a pain point every all of us have when we have spent many hours looking for the content we want to watch, can just search for, show me a romantic comedy from the 1960s with a happy ending and actually find something exactly that. So I think overall just more, more easier discovery means more streaming hours for this content and just higher licensing value for a company like Paramount. And now for me, this can be used as repurposing this content.
We have tools for repackaging this existing content into new formats as well. So for example, if you are a news media company, you will typically be under time crunch because there will be a lot of breaking news which you need to repurpose for different platforms. Now, using our Reka Reads Generation product, you can instantly create a compilation of all of the news highlights that happened in a day and just share it on different platforms, including social media platforms where a lot of your audience might be. And as a result, this generates this new audience engagement on different platforms very, very quickly and without adding any additional production costs for you. And I think one of the key things to call out here is that time is of essence in these industries. If I am a news publisher, if it’s about sports highlights, I don’t want to spend another week with content editors editing this and then sharing these in platforms. It’s kind of pointless at that point. So that’s where having, using Reka’s reuse generation product, you can simply prompt it to generate, instantly generate all of these compilations that you can share on social media platforms. Last I would also say that now as a publisher you can increase your advertising and sponsorship revenue because now you can measure your ROI much better and you can sell premium ad slots as well. So for example, if Coca-Cola is advertising with Paramount, Paramount can prove ROI much better by measuring every logo appearance, especially every logo appearance that might happen in a video. For example, in a sports studio, there might be Coca-Cola banners behind. It can now track every time
Coca-Cola is mentioned or Coca-Cola appears and then charge and prove the ROI to its advertisers. It can also sell premium placements like it can sell moments of excitement. For example, Nike can decide to only show their ads when there is a moment of excitement like a goal or something which they know is likely to make someone buy a shoe.
And this is a very premium ad placement for these publishers. And I think that’s just some of the ways, like now being able to sell these premium ad slots and just being able to prove that ROI and adding it to your sales techs are just some of the ways it can help.
Greg Kihlstrom (11:15)
Yeah, I mean, that’s that’s a lot of exciting stuff that is now now available. You know, I wonder where you’re thinking as far as, know, what are you excited about as AI evolves and, you know, over the next, let’s say, five years, you know, what do you see shifting?
Meenal Nalwaya (11:33)
a very interesting question. Five years is a lifetime. In the world of AI, I see two inevitable shifts happening. One, think, is kind of obvious one. And the second, I would say, is slightly a more Cheyenne view that I have. From the obvious one, I think the world will move from text to multimodal, which we can imagine that the first round of AI adoption is around text and coding. But then,
Greg Kihlstrom (11:35)
Right, right.
Meenal Nalwaya (12:03)
most of the world’s data or a large part of the world’s data is how humans perceive world. It’s in images, video, audio. And I think the time for that is yet to come. So I think that’s going to drive the next wave. The second is a slightly contrarian view, where I think that companies will, from rappers, we will move towards more full stack companies. Because I think as companies scale,
It would become more and more critical for them that they own the entire stack and have control over the kind of quality, reliability, or even pricing power that they have. Especially when you go deeper and start delivering complex real world use cases, it’s not going to be an ideal scenario if for every safety outcome or for every problem that you have to solve in a media company, you are dependent on someone else’s roadmap.
For example, I saw that first-hand at Meta where we were shipping multimodal in Meta AI and hallucination ended up being a big problem for us. And we realized that it was much harder to solve it at the fine-tuning layer. And we really needed to have it incorporated in the pre-training layer itself to drive that improvement. And I think that kind of control is critical.
Greg Kihlstrom (13:22)
Yeah, yeah. What advice would you have then, given some of these as well as your other experience, what advice would you have for PMs and AI?
Meenal Nalwaya (13:33)
I love this question because there is so much chatter around PMing and pretty much every job at this point in AI. First of all, think there have been some very famous statements about how PMing is going to be less relevant, et cetera, et in AI. I actually think this is the best time to be a PM because it’s much easier using tools like Cursor, et cetera, to code. And the value that PMs can really add is the what and the why, which is if you still need to define what is it that you need to do and why do you need to do that? I recently Andrew Ng actually said that I think the right PM to end ratio now is actually something like 1 PM for every two or three engineers because you can code much easier, but then you need to still know what is it that you are trying to do. But anyways, coming to the advice, so.
First, I think I have two advice that I would give for any PM in AI. One, this tech is cool, but still anchored on the customer problem. The option of your product really lives or dies there. If the AI doesn’t fit into a user’s problem or it doesn’t solve a really, really painful problem, it’s just not going to stick. It’ll just remain cool tech. And second, from an execution perspective, evals or evaluations in the world of AI at the new PID. So your product will really be defined by the success metric and that measurable outcome in this case is going to be the real customer and the real data and having very solid set of evals on what good looks like is going to make sure you’re building the product in the right direction.
