Temi Babs is a software engineer with a deep love for AI and an even deeper connection to language. He’s working to make voice technology work for African languages in a way that truly reflects how we speak, think, and live. What started as a personal struggle with speaking Yoruba has grown into Spitch, a voice AI platform helping developers build tools that actually sound like us.

In this spotlight, Temi shares how his journey began, and why he believes the future of voice tech doesn’t just include African languages, it depends on them.

We caught up with Temi to talk about how it all started.

Sure. My name is Temi, and I’m a software engineer with a focus on Artificial Intelligence (AI). I’ve been working in the AI space for about seven years now. I started out at Interswitch as a data scientist, that was right after I studied computer science at the University of Ibadan.

Since then, I’ve led engineering and machine learning teams across different industries, mostly in fintech. I’ve worked at places like Voyance, Indicina, and now, Spitch. Outside of work, I love reading, particularly scientific literature. I’m drawn to topics in physics and psychology.

I’d say my journey into AI really began at Interswitch, where I joined the research and development team. That was my first real experience working in a corporate setting. We were building all kinds of things, from deep learning models to blockchain experiments. One of the most memorable projects for me was a chatbot we built for customer support. This was before the transformer era, so we didn’t have the luxury of advanced models like ChatGPT. We used a small LSTM model to power it, but nowhere near what transformers can do today. We actually launched it internally, and the team used it for a while before it was eventually shelved. 

Then fast-forward to around 2022, I discovered OpenAI’s GPT-3 API. That was before ChatGPT became mainstream. We used it to build another chatbot. This time on WhatsApp, and it worked surprisingly well. We kept iterating, and that eventually led us to build a tool for students. It was a voice-based assistant for students. Instead of taking notes during lectures, students could just focus on listening. The assistant would transcribe everything in real time using OpenAI’s Whisper. But we ran into some challenges, especially around the transcription quality, particularly with African accents and languages. So even though we decided to pivot away from that product, we came out of it with a key insight that there was a real gap when it came to African language support in voice AI. That was the beginning of what eventually became Spitch.

Spitch came out of a pivot. But on a more personal note, it was tied to my own experiences growing up in Ibadan. My brother and I were quite sheltered as kids. We didn’t go out much or interact with a lot of other people. My parents spoke only English to us at home, and while that may sound like a privilege, it actually became a problem.

Whenever we went for family gatherings, we couldn’t really engage because we didn’t speak Yoruba well, and that made it hard to connect with our family. After university, I realized I needed to fix that. I made a conscious effort to learn Yoruba on my own. Today, I’d say I speak it with about 80% fluency, and I’m still improving. But that experience stuck with me. I kept thinking, “Why are our languages fading like this?”

There’s something powerful and painful about seeing your own language erode, especially when it limits how people express themselves or participate in society. That frustration and sense of loss was part of what pushed me to build Spitch. 

Absolutely. AI can and should play a huge role in preserving our languages. There’s so much depth, history, and expression in African languages that just isn’t represented in the digital world right now. One thing I always say is that no single person can capture the full experience of a language or culture. They’re deeply tied to how we understand the world. But beyond just preserving what’s already there, we also have an opportunity to expand our languages, to give them vocabulary for scientific terms, for things we haven’t historically expressed in them.

It’s happening already in small pockets. On Twitter, for instance, there are creators doing science in Yoruba. I remember one creator, her name is Wura, she was teaching tech in Yoruba. It’s amazing to watch. That’s where Spitch comes in. If we get this right, we could reach a point where kids grow up speaking our languages fluently and still thrive globally. They won’t need English to access tech or science or education. The tools will be there in Yoruba, Hausa, Igbo, and more. The dream is that one day, local languages won't be a limitation. They’ll be a starting point.

Right now, we see ourselves as a platform, and we’ve built it in a way that’s API-first. So, if you’re a developer or an early-stage startup and you want to build something that uses voice, you can just plug into Spitch. Let’s say you need speech-to-text or text-to-speech. Or maybe you’re building a voice AI agent, you can come to our platform, grab an API key, and start experimenting. We’ve made it as self-serve as possible so that it’s accessible not just to big enterprises, but to solo developers and young startups too.

You can go to spitch.studio, get your API key, and access our documentation. From there, you can start building. Our goal is to give people, especially in Africa, the tools to create voice-driven products affordably. We also work with some enterprises that need more custom solutions. The vision is much bigger than that. We’re heading toward a time where people and businesses can pick up our tools and build voice agents, interactive voice applications, or other products that speak and understand local languages. That’s the future we’re working toward.

Right now, we offer four core services: Speech-to-text, Text-to-speech, Prefix and Machine translation. From there, we’re building more tools, and just generally trying to improve the developer experience. 

That's a great question, and I think it's inevitable. Yes, large companies will eventually get better at this. They have the resources, the research teams and the compute. It’s just a matter of time. But the thing is that the voice landscape is huge. Way bigger than people often realize. And I think that’s where we have an edge. First, we’re much closer to the market. We understand the cultural nuances, how language actually evolves here, how people speak today versus how they spoke 10 years ago. That kind of insight is hard to replicate from outside the continent.

Take Yoruba, for instance, even the tonal structure shifts subtly over time. Vocabulary changes. Gen Zs, for example, have their own ways of speaking. So, the key isn't just building a model that works today, it's staying close to how language continues to evolve, especially locally. That’s where local companies like ours have a long-term advantage.

I also think that eventually, rather than trying to do everything themselves, these large global labs might prefer to partner with local developers like us who understand the terrain. That’s how I see it playing out.

Oh, definitely. There’ve been challenges, plenty of them. When you're building anything AI-related, you quickly realize that it takes two core things: compute and data. Compute, while expensive, is actually the easier of the two to solve. If you can raise funding, you can buy GPUs. That’s a money problem.

But data? Data is a completely different beast. The biggest issue we’ve faced is the lack of data infrastructure in this part of the world, not just the tools, but also the mindset around data. Sometimes I wonder: do people not know how valuable data is? Or do they know, but just don’t care? Or maybe they care, but aren’t positioned to do the hard work it takes to build the systems we need. You see, if you’re in the West, it’s relatively easy to find quality datasets, or to partner with providers. Here, you have to do almost everything yourself, and even then, you’re fighting uphill. There are almost no great data providers on the continent. 

I believe we need a cultural and economic shift, where data is seen as infrastructure, and people are incentivized to treat it as such. Until we do that, AI in Africa will always face bottlenecks.

It definitely hasn’t been easy. Honestly, it’s been brutal at times. But the mindset we’ve taken on as a team is pretty simple: some things just have to get done, and now is the season to eat the dirt, or as they say, “eat shit.” (Pardon my French.) Everyone on the team understands that these are the early days, the really rough ones. And that this is when we need to give it everything. So, we’re putting in crazy hours, anywhere between 12 and 16 hours a day, every day. Whether it’s research, engineering, developer experience, everyone is just doing what they can to move the needle. We know the value will come later. We believe in the mission. And for now, we’re just focused on doing the work. There’s no fancy answer to it. We're in the mud right now, but we know we’re building something meaningful. And that’s what keeps us going.

It was incredible. VivaTech really opened my eyes to how different the international stage is compared to what we’re used to here in Nigeria or even across Africa. Beyond just showcasing tech, it gave me a chance to observe human behaviour across cultures, and how they think about technology. That’s important to me because language is deeply tied to culture. So, seeing those differences helped me understand the broader context of what we’re building.

I also saw a lot of innovation happening around multimodal interfaces, combining voice, text, and visuals. AI that understands the environment it’s in, that’s aware of context, tone, surroundings, that’s where things are going. There’s also a growing emphasis on edge AI, running lightweight models on small devices without internet access.

So, for us, it confirmed that the future is diverging in two directions: Very large models that do everything and require massive infrastructure and Very small, focused models that can run offline anywhere. If you’re building in voice AI, those are the two directions to think about. And VivaTech helped me see that clearly.

Honestly, I’m driven by impact, and I measure that with real numbers. I love looking at our usage metrics, like how many hours of audio we’re processing each week? Those numbers tell a story. Behind every single hour is a person, someone actually using what we built. Seeing the range of use cases. Maybe it’s a student using Spitch to access lecture content in Yoruba. Maybe it’s a customer calling a business and getting support in Hausa.

That’s the real motivation. Seeing that this thing we’re building is actually breaking down barriers, helping people communicate, learn, access services in their own languages. So yeah, it’s hard. We’re grinding, and the road isn’t smooth. But when I see that someone, somewhere, is having an easier time because of what we built, that’s what keeps me in it.