Lelapa AI
Lelapa AI is a privately held language-technology company based in Johannesburg, South Africa. It operates as a research and product lab. Pelonomi Moiloa is CEO. The company develops African-language systems with a focus on limited computing resources and practical deployment.
Company & leadership
- Chief Executive Officer
- Pelonomi Moiloa
- Chief Technology Officer
- Jade Abbott
- Chief Information Officer
- Vukosi Marivate
Confirmed co-founders
- Pelonomi Moiloa and Vukosi Marivate; this is not a complete founder list
Origins and company focus
Lelapa's public writing was active by December 2022, when Pelonomi Moiloa published an article on responsible AI. The company's current mission focuses on language systems that work with limited compute, fragmented data and diverse language settings.
The sources reviewed confirm Pelonomi Moiloa and Vukosi Marivate as co-founders. They do not establish a complete founder list or an incorporation date, so this profile does not assign an unverified founding date.
Current leadership
Lelapa's team page lists Pelonomi Moiloa as CEO, Jade Abbott as CTO and Vukosi Marivate as CIO. Benjamin Rosman is Chief Scientist, Thembani Biyam is COO, and Mbali Ndandani is CRO. These are the roles shown on the company page checked on 6 November 2026.
Location and ownership
The company identifies Johannesburg as its base in both its company profile and privacy notice. It is privately held. Mozilla Ventures identifies Lelapa AI as a portfolio company. The reviewed sources do not disclose shareholder percentages or a complete ownership structure.
Research and commercial work
Lelapa describes itself as a research and product lab. Its work centers on efficient language systems for varied infrastructure conditions. Vulavula is its commercial transcription and translation service, while InkubaLM is part of its published model research.
Research and community milestones
In October 2024, Lelapa published an introduction to InkubaLM and its accompanying language datasets. The research received computing credits from Microsoft AI4Good.
In June 2025, Lelapa reported results from the Buzuzu-Mavi Challenge with Zindi. The model-compression project involved more than 490 participants from 61 countries. It extended the company's research into methods for running language models with fewer resources.