Ex-Goldman, Meta founders build voice AI for overlooked markets

Customer support and service are among the hottest areas in voice AI today. But building a product that sounds human and responds with minimal delay is far more challenging in some markets than in others — and most major players weren’t designed with Africa and the Middle East in mind.

AethexAI, a startup founded last year to bridge this gap, has raised $3 million in pre-seed funding led by 4DX Ventures, with participation from Enza Capital, Dorm Room Fund, Mojo Ventures, and Stanford GSB 26 Fund. Individual backers include Stanford faculty, telecom executives, and AI researchers from Anthropic.

Instead of relying on existing orchestration tools like Vapi and LiveKit, the company built its own compact model and orchestration layer from scratch to accommodate localized dialects of English, French, and Arabic spoken across its target markets — a decision driven, as we’ll see, by the specific demands of operating in the region.

The company is also launching its platform for enterprises to test its technology and sign up for services, along with APIs and SDKs for developers to experiment with its models.

The startup was founded by Mariama Diallo and Ayooluwa Odemuyiwa. CEO Diallo previously worked at Goldman Sachs and later joined YC-backed ModelML as a product and growth hire. CTO Odemuyiwa graduated from Caltech, worked at Meta, and attended Stanford Business School before co-founding the company. The pair wanted to build something for emerging markets and began looking for opportunities.

Businesses worldwide are racing to adopt AI tools to automate parts of their operations. But that doesn’t always go smoothly. In Egypt, a call center automated a large portion of its calls but reverted the system due to poor results, the founders discovered. Several support centers in Africa told them that finding and hiring engineers to automate calls at the right cost was a persistent challenge.

“The latency and jitter we saw on automated calls in this region were outrageous. If we had become orchestrators, we might have had to use large models hosted outside the region, leading to higher latency. We realized that for this to work, we need to use very small models and cut latency at every step,” Odemuyiwa told TechCrunch about the decision to build the company’s own models and orchestration layer.

AI labs that deploy their latest models typically spend millions training them and acquiring data. AethexAI found a solution for both. Instead of chasing the largest possible models, it concluded that small models are sufficient to address the latency problem while preserving accuracy, and developed its own Kora series with parameters ranging from 300 million to 1.7 billion. That’s a fraction of the size of large language models — and that’s precisely the point.

To train these models, the startup used anonymized recordings from a call center partner. It also shipped hard drives to radio stations across Africa to collect more audio data. To keep costs low, it built a contributor network of university students to annotate data and pronounce local names. As a result, the startup says it now handles more than 17,000 calls per day.

On the business side, the company is careful to guide clients who are new to voice AI through the process, offering onsite demos and workshops to help them identify the best use cases for automation.

“We always tell customers that we can’t be everything for everyone right now. We’re small. When we start talking to a company, we ask them to pick one use case that matters most to them to begin with,” Diallo said.

The startup is open to working across all industries, but at the moment a large portion of its use cases involves calls for debt collection, customer activation, or KYC — Know Your Customer verification, the standard identity-checking process used by banks and telecoms. The company is hiring forward-deployed engineers on a contract basis to serve local markets and building channel partnerships with telecom providers to handle telephony for voice AI calls. Plug-and-play solutions, it says, simply won’t work here.

Walter Badoo, co-founder and managing partner of 4DX Ventures, argues that the Africa and Middle East market is fundamentally different from the markets most voice AI companies were built to serve.

“Enterprises in Africa and the Middle East process roughly three times the call volume of their Western counterparts, as voice is still the dominant channel for customer interaction,” he said. “Incumbent systems were built for Western markets characterized by high-end GPU infrastructure, standard English and European speech environments, and enterprise workflows common in the US and Europe. That creates real gaps when enterprises need systems that handle dialects, code-switching, and informal speech patterns, and that work within their existing telephony infrastructure and their actual price points.”

Put another way, while companies like ElevenLabs, Deepgram, Sierra, and Cognigy are expanding globally at a fast pace, the markets they were built for and the markets they are entering aren’t always the same. Startups like AethexAI are betting that the gaps — models specialized in local dialects, on-the-ground partnerships, infrastructure built for the region — represent a market opening that the giants have neither the incentive nor the architecture to close.