AethexAI: The Voice AI Targeting Africa and the Middle East

A Voice AI for Overlooked Markets

In the voice AI sector, customer support and service are rapidly evolving fields. However, creating a product that sounds human and responds without noticeable delay is particularly challenging in certain markets, notably in Africa and the Middle East. Large companies often do not design their solutions with these specific regions in mind.

AethexAI, a startup founded last year, has set out to fill this gap. It recently 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. Among the individual investors are Stanford professors, telecommunications executives, and AI researchers from Anthropic.

A Tailored Approach

Unlike other companies that use existing orchestration tools like Vapi and LiveKit, AethexAI has chosen to develop its own models and orchestration layer. This decision was driven by the specific requirements of the region, including the management of localized dialects of English, French, and Arabic. Rather than relying on large models hosted outside the region, which would lead to increased latency, AethexAI opted for very small models to reduce latency at every stage.

The startup also offers a platform for businesses to test its technology and sign up for its services, as well as APIs and SDKs for developers. This allows companies to immerse themselves in the technology and discover its potential applications.

Visionary Founders

AethexAI was founded by Mariama Diallo and Ayooluwa Odemuyiwa. Diallo, who worked at Goldman Sachs, was also a product and growth lead at ModelML, backed by Y Combinator. Odemuyiwa, a former Meta employee and Caltech graduate, enrolled in Stanford's business school before co-founding the company. The duo aimed to create a solution for emerging markets and began exploring opportunities in these often-overlooked regions by large companies.

Addressing Local Needs

Companies worldwide are looking to adopt AI tools to automate some of their operations. However, this does not always work. In Egypt, for example, a call center had to revert to manual methods after disappointing results with automation. The founders of AethexAI found that latency and jitter were major issues in the region. Several support centers in Africa reported that finding and recruiting engineers to automate calls at a reasonable cost was a real headache.

To tackle these challenges, AethexAI developed small models, reducing latency while maintaining accuracy. The Kora series, with parameters ranging from 300 million to 1.7 billion, is designed to be efficient under these conditions. This is a fraction of the size of large language models, which is precisely the goal.

Innovative Data Collection

To train its models, AethexAI used anonymized recordings from a call center partner. It also shipped hard drives to radio stations across Africa to collect more audio data. To cut costs, it built a network of contributors made up of university students to annotate the data and pronounce local names. Thanks to these efforts, it now handles over 17,000 calls per day.

A Targeted Business Strategy

AethexAI helps its clients discover voice AI, offering on-site demonstrations and workshops to assist them in identifying the best use cases for automation. The company is open to working across all sectors, but for now, a significant portion of its use cases involves calls for debt collection, customer activation, or identity verification (KYC). The company is hiring field-deployed engineers on contract to serve local markets and is establishing partnerships with telecommunications providers to manage the telephony of voice AI calls. Off-the-shelf solutions simply will not work here.

A Fundamentally Different Market

Walter Baddoo, co-founder of 4DX Ventures, emphasizes that the African and Middle Eastern markets are very different from Western markets. Companies in these regions handle about three times more calls, as voice remains the primary channel for customer interaction. Existing systems were designed for Western markets characterized by high-end GPU infrastructure, standard English and European speech environments, and common enterprise workflows in the U.S. and Europe. This creates real gaps when companies need systems capable of handling dialects, code-switching, and informal speech patterns, and that work within their existing phone infrastructure and true price points.

In conclusion, AethexAI is tackling a niche that large voice AI companies have yet to explore, betting on specialized models and infrastructure tailored to emerging markets.