Zimbabwean entrepreneurs Michael Moyo and Louis Murerwa have raised US$2 million in pre-seed funding for Ocular AI, a company that develops datasets and evalua...
Zimbabwean entrepreneurs Michael Moyo and Louis Murerwa have raised US$2 million in pre-seed funding for Ocular AI, a company that develops datasets and evaluation tools for advanced artificial intelligence systems.
The funding round was led by Drive Capital, with participation from Y Combinator, Alumni Ventures, 1745 Ventures, Orange Collective, MyAsia VC and angel investors.
The company says it has already reached seven-figure revenue and supplies data to leading AI research laboratories and Fortune 100 companies, although it has not publicly named its customers or disclosed an exact revenue figure.
Ocular AI's business model places it in the AI industry's data supply chain, providing material used to train and evaluate models rather than selling consumer-facing AI applications.
The company announced the investment on 5 October 2026, saying the money would support its work building an applied AI data research laboratory.
What Ocular AI does
Founded in 2024, Ocular AI develops datasets, benchmarks and evaluation tools used to assess the performance of advanced AI models. Its work initially focused on voice technology, with plans to expand further into systems that combine audio and video.
The company’s products are designed to help AI developers identify weaknesses that may not be apparent through conventional testing. These include difficulties understanding accents, recognising pauses and interruptions, and responding naturally when two people speak over each other.
Traditional voice systems often convert spoken words into text before processing a request and generating a spoken response. This approach can lose information carried by tone, timing, emphasis and other features of natural conversation.
Ocular AI is developing datasets that preserve more of these features. Its full-duplex conversational data records two speakers separately, allowing researchers to study how speech overlaps and how people respond to each other in real time.
Growing demand for voice-native AI
The company’s work comes as technology developers pursue AI systems that can listen and speak simultaneously, rather than treating conversations as separate stages of transcription, text processing and speech generation.
Such systems need examples of real conversations to learn how people communicate, including interruptions, hesitation and changes in speaking pace. Ocular AI argues that high-quality conversational data is difficult to obtain from publicly available online material.
In a research article published in May 2026, the company highlighted the continued use of an older telephone conversation dataset recorded in 2004. It argued that the industry needs higher-fidelity recordings to support newer voice models.
Ocular AI has also developed the Converse benchmark family to evaluate voice models. Its Converse-STT benchmark tests how accurately speech-to-text systems transcribe real conversations, including speech patterns that may be absent from clean, scripted recordings. The company published the benchmark in September 2026.
The startup is now extending its work to audiovisual AI. In September, it introduced datasets containing synchronised audio and video from two-person conversations, allowing researchers to examine speech alongside facial movement and other visual cues.
Expert network supports dataset development
Ocular AI says its work relies on a network of vetted specialists who contribute professional knowledge to training data and evaluations.
The company’s website describes a network of more than 10,000 domain experts across more than 40 languages and 120 disciplines. Its specialists include professionals such as doctors, lawyers, engineers and linguists.
The approach combines specialist knowledge with research into model performance. Experts help establish what an accurate or appropriate response should look like, while evaluations identify where an AI system falls short. The findings can then inform new datasets and testing methods.
This process is intended to give developers more reliable ways to assess whether their models can handle practical tasks, rather than relying only on standardised tests.
Funding and business details
Drive Capital led Ocular AI’s US$2 million pre-seed round. Other participants included Y Combinator, Alumni Ventures, 1745 Ventures, Orange Collective, MyAsia VC and angel investors, according to the company’s announcement. 1745 Ventures was previously known as Bertelsmann Digital Media Investments.
Ocular AI reports that its revenue has reached seven figures and is growing, and that major AI laboratories and Fortune 100 companies use its datasets and evaluation tools. These figures and customer claims are company-reported; the company has not publicly identified those customers or disclosed an exact revenue total.
The startup is based in San Francisco, California, according to its website. While the founders have connections to Zimbabwe and the wider region, coverage of the company identifies Michael Moyo as Zambian and Louis Murerwa as having grown up in Zimbabwe. It is therefore more precise to describe Ocular AI as a company co-founded by entrepreneurs with Zimbabwean and Zambian connections, rather than identifying both founders as Zimbabwean.
The latest funding gives Ocular AI additional capital as it expands its data products and evaluation tools for advanced AI systems. Its progress will depend in part on whether demand for specialist datasets and realistic conversational benchmarks continues to grow as AI companies develop more capable voice and audiovisual models.
