The world of voice AI is buzzing, particularly in customer support and service. However, crafting a truly human-sounding, low-latency AI solution proves remarkably challenging in certain regions.
Many prominent voice AI platforms simply weren’t designed with the unique linguistic and infrastructural demands of markets like Africa and the Middle East in mind. This significant gap is precisely what AethexAI, a startup founded last year, aims to bridge.
The company has successfully secured $3 million in pre-seed funding to fuel its mission. This round was spearheaded by 4DX Ventures, with additional contributions from Enza Capital, Dorm Room Fund, Mojo Ventures, and the Stanford GSB 26 Fund.
Impressively, individual investors include Stanford faculty members, seasoned telecom executives, and cutting-edge AI researchers from Anthropic. This diverse backing underscores the potential and innovative approach AethexAI brings to the table.
Visionary Founders Tackling Overlooked Markets
AethexAI was co-founded by two visionary individuals, Mariama Diallo and Ayooluwa Odemuyiwa, who both left high-profile roles to pursue this venture. CEO Mariama Diallo previously honed her skills at Goldman Sachs before joining YC-backed ModelML in a key product and growth position.
CTO Ayooluwa Odemuyiwa is an alumnus of Caltech, with experience at Meta, and was enrolled at Stanford Business School before embarking on this entrepreneurial journey. The duo shared a common desire to innovate for emerging markets, and their search led them directly to the voice AI opportunity.
Globally, businesses are eagerly adopting AI tools to streamline operations, but this isn’t always a smooth transition, especially in diverse regions. For instance, the founders discovered an Egyptian call center that had to roll back its automated system due to abysmal performance.
Furthermore, numerous support centers across Africa shared a persistent challenge: finding and affording skilled engineers to implement effective call automation. The core issue often boiled down to the limitations of existing voice AI in these specific environments.
A Ground-Up Approach to Latency and Localization
One of the biggest hurdles AethexAI identified was the unacceptable latency and ‘jitter’ experienced on automated calls in the region. CTO Odemuyiwa emphasized that relying on large models hosted outside the region would only exacerbate these issues, leading to higher delays.
This critical insight led AethexAI to a bold decision: to forgo existing orchestration tools like Vapi and LiveKit, and instead build its own small models and orchestration layer from scratch. This custom approach ensures minimal latency and optimal performance tailored to local conditions.
Their solution focuses on handling the intricate, localized dialects of English, French, and Arabic prevalent across their target markets. This deep understanding of regional speech patterns, including code-switching and informal language, is crucial for natural interactions.
Rather than chasing the largest possible models, AethexAI developed its proprietary Kora series, featuring models with parameters ranging from 300 million to 1.7 billion. This fractional size compared to massive LLMs is a deliberate strategy to maintain accuracy while drastically cutting latency.
To train these specialized models, the startup employed an ingenious data collection strategy. They utilized anonymized recordings from a call center partner and even shipped hard drives to radio stations across Africa to gather more diverse audio data.
To ensure cost-effectiveness and local relevance, AethexAI built a contributor network of university students to annotate data and accurately pronounce local names. This meticulous, localized approach has paid off, with the platform now successfully handling over 17,000 calls per day.
Driving Impact and Sustainable Growth
AethexAI isn’t just delivering cutting-edge technology; they’re also committed to guiding their enterprise clients through the voice AI adoption process. They offer onsite demos and interactive workshops to help businesses identify the most impactful use cases for automation.
“We always tell customers that we cannot be everything for everybody right now. We’re small,” explains CEO Mariama Diallo. “When we start talking to a company, we ask them to pick one use case that is the most important to them to start [with].”
While open to all industries, AethexAI is currently making significant strides in areas like debt collection, customer activation, and Know Your Customer (KYC) verification for banks and telecoms. These are high-volume, critical processes where efficient voice AI can make a huge difference.
To effectively serve these diverse local markets, the company is strategically hiring forward-deployed engineers on a contract basis. They are also forging vital channel partnerships with telecom providers to seamlessly integrate telephony services for their voice AI calls, understanding that “plug-and-play” solutions simply won’t suffice here.
The Untapped Potential of Underserved Markets
Walter Badoo, co-founder and managing partner of 4DX Ventures, eloquently articulates the profound differences in the Africa and Middle East market. He highlights that enterprises in these regions process approximately three times the call volume compared to their Western counterparts, as voice remains the primary customer interaction channel.
Incumbent voice AI systems were designed for Western markets, which typically feature high-end GPU infrastructure, standard English and European speech environments, and established US/European enterprise workflows. This creates substantial gaps when local enterprises require systems capable of handling:
- Diverse dialects
- Code-switching
- Informal speech patterns
- Integration with existing telephony infrastructure
- Cost-effective pricing points
While global giants like ElevenLabs, Deepgram, Sierra, and Cognigy expand rapidly, AethexAI recognizes that their core architectures and incentives aren’t built for these specific challenges. The startup is strategically leveraging these critical gaps—specialized local models, on-the-ground partnerships, and regionally optimized infrastructure—as a unique market opening.
AethexAI is making a bold bet that the future of voice AI in these vibrant, high-growth markets belongs to those who understand and build for their distinct needs, proving that innovation can thrive by looking beyond the obvious.
Source: TechCrunch – AI
