Building the AI voice for global storytelling

Most MENA startup media barely scratches the surface. Our thesis is to go deeper. We dig into the why, the how, the strategy, the implications, and the hard questions, extracting insights you can use to build better.

Members get access to thoughtful analysis, case studies, playbooks, and interviews that unpack the most important companies, trends, and shifts in MENA’s startup, tech, and VC landscape.

🔓 Don’t miss out! Grab a FREE 7-day trial to unlock full access and try it for yourself.

Members include people at BECO Capital, Instabug, Antler, Carta, ADNOC, Plug and Play and more, so you’ll be in good company!

Hi friends! 👋

Last week, the eyes of the world turned to the Gulf.

Between Donald Trump’s whirlwind visit, the lifting of NVIDIA chip restrictions, a deluge of AI dealmaking, and billions pouring into data centres, the Gulf’s AI ambitions collided with geopolitics in what felt like a Baz Luhrmann–style fever dream.

There will be ample time for us to unpack all of that in the coming weeks, but for now, we want to bring things back to a discourse that really started gaining traction around the time of the DeepSeek moment at the turn of the year, when we were suddenly inundated with everyone’s two cents on when MENA was going to have its regional AI champion and day in the sun.

But, what if we already have one?

Tucked inside a modest office in Dubai, a father-son duo is quietly building one of the world’s most advanced real-time voice localisation platforms. Not in Silicon Valley. Not in Shenzhen. Right here.

Their company CAMB.ai doesn't just simply clone voices; it also preserves emotional fidelity, synchronises lip movement, and translates live sports commentary across dozens of languages in real-time.

Founded by Akshat Prakash (CTO), an engineer who previously worked on the Siri/Apple AI/ML team led by GANs inventor Ian Goodfellow, and his father, Avneesh (CEO), a veteran B2B operator, CAMB.ai was born from a deeply personal experience: the alienation of language.

The duo were highly conscious from the off that they didn’t want to build yet another text-to-speech tool. What mattered to them was building storytelling infrastructure; preserving tone, emotion, and intent. And they've quietly been doing just that.

Today, CAMB.ai powers real-time multilingual broadcasts for the MLS, the Australian Open, and more. Their speech engine, MARS, now in its seventh iteration, was the first generative speech model deployed on Google’s Vertex AI and Amazon’s Bedrock. Go to the Model Garden and you’ll find them listed right alongside Meta, Anthropic, HuggingFace, and Mistral.

In March 2025, the company closed an $11M pre-Series A, bringing its total raised to $15.5M, with backing from a global slate of sports-tech VCs and strategic media players.

In this conversation, Akshat and I dig into what it takes to build foundational AI when your ecosystem is still catching up, how you earn deep technical defensibility by actually understanding domain nuance, and why the Middle East doesn’t need more capital, it needs more belief.

We also get into why localisation is the hardest problem in generative AI, and why live sports, with all its chaos and emotion, was the ultimate training ground.

Here’s what we covered:

• 🎙️ Why real-time speech localisation is a systems problem, not just a model problem

• 🧱 How building your own streaming infra unlocks product edge in high-latency environments

• 🏋️‍♂️ Why CAMB built a “sports mentality” company, and why being second-best is a superpower

• 🔓 How open-sourcing Mars 5 turned skeptics into believers, and became a hiring magnet

• 🚀 Re-writing the global perception of where great AI can be built

Let’s get into it 👇

Actionable insights 🧠 🛠️

If you're short on time, here’s what the smartest builders, funders, and infrastructure-layer founders can take from CAMB.ai's trajectory – and why it matters for anyone building in AI, frontier markets, or real-time media infrastructure.

Premium members get the full version of this article, plus a TLDR summary with key takeaways and actionable insights right here.

Okay, let’s dig into it 👇

You co-founded CAMB.AI with your father, which needless to say, is a rare family dynamic. I’m curious about the complementary skill sets you both bring to the table, and also about that bigger backdrop. What was the “this is it” moment? Why did this feel like the problem you were going to build together?

Yeah, I think as you said, we have really complementary skill sets. My father has been in the industry for 35 years, building large B2B companies. He’s a proper business CEO. I’ve been in the tech world for far less time, but in some fairly significant places — at Siri, for example. I’ve been building this tech for a while now, especially when you consider that the entire lifetime of GenAI is maybe two and a half years. I’d say I’ve spent at least three to four times that amount of time preparing for it — doing the work behind the curtain, so to speak, before the world really saw it.

So yes, we complement each other well. But more than just skills, the problem we’re solving has always felt personal. And there’s a bit of irony in that — because my dad and I are 30 years apart in age, and the world of technology has completely changed in that time. Today, I can click a button to order a cab, food, clothes, or book a movie — all instantly. Thirty years ago, he couldn’t do any of that. But language? Language is still the same problem.

The way he experienced it is the way I did. When I went to Carnegie Mellon, all my tests were in English. Even though I’d spoken English all my life, I still felt uncomfortable understanding what the questions actually meant. Or something as simple as ordering coffee at Starbucks — if someone said my name or order in a super-fast American accent, I’d be thinking, “Jeez, where are the subtitles?”

Even when you speak the same language, just the accent can be enough of a barrier. That shared experience — of being alienated by language — is what brought us together. We knew we wanted to build a great company together. The idea just became the excuse to do it.

And when we really introspected, this was the most meaningful thing we could pursue. It was a deeply shared experience, and so culturally relevant — and we realised the kind of impact we could make. Most of our business ventures, especially for my father, have been around inclusion — whether it’s financial inclusion, identity inclusion... he’s led massive global projects focused on including more people.

This is just another version of that. For me, it’s a perfect amalgamation of the speech technology and speech AI I’ve been working on, and that shared problem we both faced. So it wasn’t like, “We entered this space and voilà, here’s the idea!” It was more like: we wanted to build something great together. And this was the most meaningful thing we could’ve spent our time on.

I also want to ask you about timing, because as you mentioned, you and your father experienced the same problem 30 years apart. That suggests something significant about now. So where did the bar need to be, technically or architecturally, before CAMB.ai’s core product didn’t just feel desirable, but actually felt feasible?

I think honestly — it’s a little bit of luck. That plays a huge part. If we had started exploring this problem five years ago, I would’ve been a sophomore in college who didn’t know squat about speech tech. So timing really matters.

But one of the things I was blessed with is that during my time at Siri, I worked in a group led by Ian Goodfellow.

You might know the name — he’s the inventor of GANs, the tech behind deepfakes. Long before Midjourney or Stability, or even ChatGPT, there was a moment on social media when every other video was someone pasting Tom Cruise’s face on top of theirs. That’s GANs — generative adversarial networks — and Ian invented that technology. He led the AI org at Siri when I joined.

The remainder of this interview is for members.

Go beyond the news with analysis, case studies, playbooks and interviews that will make you better at what you do.