These two founders left Goldman and Meta to build voice AI for markets that everyone ignored

Customer support and service are among the hottest sectors in voice AI right now. But building a product that feels human and responds without noticeable delay has proven to be much harder in some markets than others — and most of the major players weren’t built with Africa and the Middle East in mind.

AethexAIa startup founded last year to fill this gap, has raised $3 million in seed funding led by 4DX Ventures, with participation from Enza Capital, Dorm Room Fund, Mojo Ventures, and Stanford GSB 26 Fund. Individual investors include Stanford faculty, communications executives, and artificial intelligence researchers from Anthropic.

Instead of using existing formatting tools such as Coat of arms and Live Kitthe company built its own micromodel and coordination layer from scratch to handle local dialects of English, French, and Arabic spoken across its target markets — a decision driven, as we’ll get to, by the specific requirements of operating in the region.

The company is also launching its platform for enterprises to try out its technology and subscribe to its services, along with application programming interfaces (APIs) and software development kits (SDKs) for developers to try out its models.

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

Companies around the world are racing to adopt AI tools to automate parts of their operations. But this doesn’t always work. In Egypt, the founders of a call center discovered that a large portion of its calls were automated, but retreated from operating the system due to poor results. Many support centers in Africa told them that finding and hiring engineers to automate calls at the right cost was a constant headache.

“The latency and confusion we saw on robocalls in this region was outrageous. If we had become an orchestrator, we probably would have had to use large forms that were hosted outside of the region, resulting in higher response times. We realized that for this to work, we had to use very small forms and reduce latency at every step,” Odemuyiwa told TechCrunch of the decision to build the company’s forms and orchestration layer on its own.

AI labs that deploy their latest models typically spend millions to train them and acquire data. AethexAI has found a solution for both. Instead of going after the largest possible models, it decided that small models were enough to address the latency problem while maintaining accuracy and developed its own Kora series, with parameters ranging from 300 million to 1.7 billion. This is a small fraction of the size of LLMs, and that’s precisely the point.

To train these models, the startup used anonymous 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, she built a network of college student contributors to annotate statements and pronounce local names. As a result, the startup says, it now handles more than 17,000 calls a day.

On the business side, the company is keen to guide clients new to AI expression through the process, offering on-site demos and workshops to help them identify the best use cases for automation.

“We always tell clients that we can’t be everything to everyone right now,” Diallo said. “We are a small company. When we start talking to a company, we ask them to choose one use case that is most important for them to start (with).”

The startup is open for business across all industries, but right now, a large portion of its use cases involve debt collection calls, customer activation, or KYC – Know Your Customer verification, which is the standard identity verification process used by banks and telcos. 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 AI-powered voice calls. Plug-and-play solutions simply won’t work here, she says.

Walter Bado, co-founder and managing partner of 4DX Ventures, says the African and Middle Eastern market is fundamentally different from the markets most voice AI companies are set up to serve.

“Businesses in Africa and the Middle East handle nearly three times the volume of calls as their Western counterparts, as voice remains the dominant channel for interacting with customers,” he said. “Existing systems are built for Western markets that feature sophisticated GPU infrastructure, standard English and European speech environments, and enterprise workflows common in the US and Europe. This creates real gaps when organizations need systems that handle accents, code-switching, and informal speech patterns, and that work within existing telephony infrastructure and physical price points.”

In other words, while companies like ElevenLabs, Deepgram, Sierra, and Cognigy are expanding globally at a rapid pace, the markets they were built for and the markets they enter are not always the same. Startups like AethexAI are betting that the gaps — specialized models in local dialects, on-the-ground partnerships, and infrastructure tailored to the region — represent the opening of a market that giants have neither the incentive nor the structure to close.