Undergraduate Researcher NLP · ML · AI engineering Champaign, IL

Adaora
Mbanefo

adore-ah

Hi, I'm Adaora! I'm an undergraduate studying statistics, linguistics, and actuarial science with a minor in computer science. I'm into NLP, machine learning, and AI engineering. Outside the lab I read fiction, play electric guitar and squash, and keep a running list of words I love in languages I'm still learning.

· · · · · · · · · · · · · · · · · · · · · · · · · · · ·

Interests

What I'm thinking about

My active research lives in NLP and computational linguistics. I work with deep learning architectures (LSTMs, transformer-based models like BERT) and eye-tracking data to examine how teens and adults process text, and whether computational models can predict their eye movements from visual and linguistic features. But my interests run wider than what I'm researching at any given moment.

I want to work in machine learning and AI engineering more broadly. Not just studying models but building them, shipping them, and figuring out why they break. I'm also curious about quant trading.

Topics

NLP Machine learning AI engineering Quant trading Transformers Statistics Cognitive modelling Currently obsessed with Multi-party dialogue

Why this matters

Africa shouldn't be left behind

All of this connects to something bigger I think about a lot. The technical gap between the African continent and the rest of the world is widening fast, especially in AI and language technology. But the gap isn't about talent. Africa has brilliant people, researchers, and builders. What's missing is the infrastructure and the accessibility.

And accessibility is doing a lot of work in that sentence. It means students in higher education without reliable access to a computer. It means networks that drop in and out. It means someone who is curious and motivated and ready to learn, but hits a wall because every tutorial, every model, every interface assumes a language they don't speak. These aren't edge cases. They're the everyday reality of millions of people the technology was supposedly built to help.

Generative AI is one of the most powerful resources of our time, and every person should have the ability to use it and understand it! So part of why I do this work is to help bridge that gap. Not by deciding from the outside what Africa needs, but by supporting the people already building, and by working on tools that meet people where they actually are.

One thread I'm especially interested in: most of the AI being deployed today is trained largely on Western data, which means it doesn't really understand African contexts — socioeconomic realities, living situations, food, family structures, the texture of daily life. I want to work on models that aren't just usable by Africans but actually understand them. An Africa-centred LLM is a project I'm planning to explore.

Closer to home, I'm also looking at ways to improve life for people back home in South Africa through initiatives focused on the youth. More on that as those projects take shape; a few live on the projects page.

Where I'm focused

  • →  Access to AI tools, for anyone who wants them
  • →  Models that understand African contexts, not just include them
  • →  Research that incorporates African languages
  • →  Youth-focused initiatives in South Africa
  • →  On-ramps for African students into NLP & AI
  • →  More to come. This is a long game.

Currently

What's on the desk

Working on
A paper on transformer-based models of reading behaviour, plus speech analysis at UChicago's ChatterLab.
Reading
Tahereh Mafi's This Woven Kingdom series.
Learning
Speaker diarisation, and Mandarin.
Listening
Dominic Fike (always).

Selected work

A few things I'm proud of

Research · ongoing

Understanding Reading Development through Visual and Linguistic Factors

Examining whether deep learning architectures (LSTMs, transformer-based models like BERT) can model and predict the eye movements of teens and adults from visual and linguistic features of text.

Research · current

Naturalistic speech, classified

Automated classification of multi-party conversation in real home environments. A large speech corpus, PyTorch pipelines, and a lot of interesting failure modes.

Project · ongoing

Littafin Fasaha

An interactive computer science learning platform for Hausa speakers, with the goal of extending to other underrepresented African languages. LLM fine-tuning and frontend work; currently in the second phase, building out the platform itself.