Hi there 👋! My name is Everlyn Asiko Chimoto. I am currently a research scientist👩🏿💻 at Lelapa AI. I recently completed my PhD in Applied Mathematics at the University of Cape Town, supervised by Prof. Bruce Bassett.
My research focuses on developing data-efficient and model-efficient techniques for low-resource languages. Broadly, I study how to extract maximal performance from limited data and limited budgets.
My work spans:
- Active learning and data selection for machine translation and automatic speech recognition (ASR) models
- Model compression and quantization for NLP and ASR
- Code-switched automatic speech recognition (ASR) for low-resource languages
- Large language models (LLMs) for low-resource settings
In my PhD, I focused on improving translation tools for African languages that don’t have much digital data. Instead of asking for more data (which is expensive and scarce),
I asked: what if we use the little data we have more intelligently? I developed methods that identify the most useful examples, remove wasteful ones, and strategically select new sentences to translate (active learning).
The result: stronger translation models built with far less data and lower cost.
I hold an MSc in Mathematical Sciences from the African Institute for Mathematical Sciences (AIMS) and the University of Cape Town (UCT), where my research also centered on machine translation for low-resource languages😄. Prior to that, I earned a BSc in Computer Science from Masinde Muliro University of Science and Technology.
Beyond research, I actively contribute to growing the AI community in Africa👩🏿🏫.
I organize workshops such as AfricaNLP and NLP@Indaba, have mentor students through Moringa School, KamiLimu and NeuroMatch, and support early-career researchers entering multilingual NLP(Masakhane and KenyaNLP).
CV (PDF)»
