These two founders left Goldman and Meta to build voice AI for markets everyone else overlooked

Customer support and service present a significant opportunity for voice AI, yet many major players overlook the unique challenges of markets like Africa and the Middle East. These regions often contend with specific linguistic nuances and varying infrastructure capabilities. This gap inspired the creation of AethexAI, a startup aiming to provide tailored voice AI solutions for these underserved areas.

Founded by Mariama Diallo, formerly of Goldman Sachs, and Ayooluwa Odemuyiwa, who worked at Meta, AethexAI has successfully raised $3 million in pre-seed funding. This capital infusion, led by 4DX Ventures, with additional participation from Enza Capital and others, will fuel the company’s mission to develop voice AI specifically designed for emerging markets.

Rather than relying on existing orchestration tools, AethexAI has built its own small AI models and orchestration layer from scratch. This strategic decision was driven by the necessity to handle localized dialects of English, French, and Arabic efficiently, thereby overcoming the latency and jitter issues prevalent in these regions. The company emphasizes the use of small models to minimize latency and ensure optimal performance.

To train its proprietary "Kora" series of models, which range from 300 million to 1.7 billion parameters, AethexAI utilizes anonymized call center recordings and collected audio data from radio stations across Africa. A contributor network of university students assists in data annotation and pronunciation of local names, ensuring linguistic accuracy and cost-effectiveness. This approach enables the startup to process over 17,000 calls daily.

AethexAI is actively engaging enterprises by offering onsite demos and workshops to help them integrate voice AI into their operations, focusing on specific high-impact use cases like debt collection, customer activation, and Know Your Customer (KYC) verification. The company also fosters partnerships with telecom providers and hires forward-deployed engineers to cater to local market demands, acknowledging that generic plug-and-play solutions are ineffective in these diverse environments.

Walter Baddoo of 4DX Ventures highlights that African and Middle Eastern markets process significantly higher call volumes than their Western counterparts, with voice remaining the primary customer interaction channel. Existing voice AI systems, built for Western contexts with different infrastructural and linguistic norms, fail to adequately address the dialects, code-switching, and informal speech patterns common in these regions. AethexAI