AethexAI, a voice AI startup targeting the African and Middle Eastern markets, has raised $3 million in pre-seed funding and launched an enterprise trial platform, API, and SDK, attempting to enter the local customer service and call automation market.
The funding round was led by 4DX Ventures.
This funding round was led by 4DX Ventures, with participation from Enza Capital, Dorm Room Fund, Mojo Ventures, and Stanford GSB 26 Fund. Individual investors include Stanford faculty, telecommunications industry executives, and AI researchers from Anthropic.
AethexAI was founded last year by Mariama Diallo and Ayooluwa Odemuyiwa. Diallo previously worked at Goldman Sachs before joining Y Combinator's ModelML team, where she was responsible for product and growth. Odemuyiwa graduated from Caltech, worked at Meta, and then went on to Stanford Graduate School of Business.
Self-developed small model reduces call latency
Instead of using off-the-shelf orchestration tools such as Vapi and LiveKit, the company built its own small models and orchestration layers, focusing on handling the local dialects of English, French, and Arabic commonly used in the target market.
The two founders stated that their research revealed some African and Middle Eastern companies that attempted to automate their call centers but reverted to manual processes due to unsatisfactory results. Other companies faced difficulties in recruiting automation engineers and controlling costs.
AethexAI believes that one of the core challenges of local voice AI is latency. Relying on large models deployed outside the region results in more noticeable latency and jitter during calls. Therefore, the company has opted to use smaller models to shorten response times at each stage.
More than 17,000 daily calls have been processed.
The company has developed its own Kora series of models, with parameter sizes ranging from 300 million to 1.7 billion, significantly smaller than mainstream large language models. AethexAI believes that this scale is more suitable for balancing speed and accuracy in the target market.
For data collection, the company uses anonymous recordings from its call center partners and also sends hard drives to radio stations in various parts of Africa to collect more audio data. Simultaneously, the company organizes university students to participate in data annotation and local name pronunciation compilation to reduce training costs.
AethexAI states that its system currently processes over 17,000 calls per day. Current primary applications include debt collection, customer activation, and KYC (Know Your Customer) verification, common in the banking and telecommunications industries.
Start by targeting enterprise clients in a single scenario.
In terms of commercialization, the company helps businesses select suitable scenarios for automation through live demonstrations and workshops, rather than covering all processes at once. The founding team stated that they currently ask customers to first select the most important scenario to start deployment.
The company is also hiring local engineers on contract and establishing channel partnerships with telecom operators to handle the telephone infrastructure needed for voice AI calls. AethexAI believes that directly replicating plug-and-play solutions from the European and American markets will be difficult to adapt to local network, language, and cost conditions.
The lead investor, 4DX Ventures, stated that voice interaction volume in African and Middle Eastern companies remains significantly higher than in Western markets. Local businesses need systems that can handle dialects, mixed codes, and informal expressions, which leaves room for regional voice AI companies to grow.
