I am working across AI engineering, data science, and machine learning — my work moves toward problems where technical rigor reaches communities current technology overlooks.
NLP for low-resource African languages · Edge AI and embedded systems for African infrastructure · Narrowing the digital divide through quiet, deliberate work.
A Luganda saying — so little done, so much more to do. Three projects already shipped, with the rest of the road still ahead.
Luganda is spoken by millions across Uganda. It has almost no public NLP infrastructure — no labeled corpus large enough to train against, no off-the-shelf classifier to fall back on. The task was not only to fit a model, but to make one possible.
I assembled a 2,600-article corpus by combining native sources, careful machine translation of English news, and synthetic augmentation calibrated to the language's morphology. I fine-tuned Multilingual BERT, then served the model through a small FastAPI layer with a lightweight web demo so anyone could try it on a paragraph of their own.
The classifier reaches 85.7% accuracy across five categories. The corpus, more than the model, is the part I expect to outlive this project.
Python · PyTorch · Hugging Face · FastAPI
Operators in Uganda, Cameroon, Malawi, and Lesotho were comparing performance against the WHO/UNICEF service ladder by spreadsheet. Regulators could not run their own questions; engineers became a bottleneck for everyone else.
I designed a role-based dashboard separating regulator, operator, and admin views over one schema, and built service-ladder visualizations that drill down by country. A small assistant, backed by Gemini, translates plain-language questions into pandas queries against the live data, with previews and guard-rails before anything runs.
Shipped during the Equitech Futures Fellowship as production tooling for partner utilities. People without engineering backgrounds can ask their own questions and get answers they can cite.
Python · Streamlit · Plotly · Gemini API · pandas
Medical-imaging models can learn the wrong things — backgrounds, scanner artifacts, label leakage — and still post good numbers in the lab. The failures show up later, in places the original authors do not see.
I am probing shortcut learning in malaria-detection models trained on Ghana health data, comparing what the network actually attends to against the regions a clinician would. The work is ongoing.
Early-stage, but oriented toward a deployment standard, not a benchmark score: a model that can be honestly used in a clinic in Accra, not only one that ranks well on a held-out set.
Python · PyTorch
Computer Science at Uganda Christian University. Equitech Futures Applied Data Institute Fellow. The work moves between NLP research on low-resource languages, data systems for social impact, embedded engineering, and leadership across technical and operational rooms. The practical question underneath it all is the same one, what does it take to deploy something useful where infrastructure is thin?
Applied Data Institute · 1 of 25 selected globally
A 10-week intensive in Bayesian statistics, deep learning, and causal inference — running alongside data work for water utilities in Uganda, Cameroon, Malawi, and Lesotho, and a parallel research in investigating spurious correlation thread on malaria detection models.
Where curiosity led
Conferences, workshops, and convenings — including Stanford's Precision Medicine & Data Science in Diabetes program. Time spent listening to where the field is heading, what others are wrestling with, and which questions deserve to come home with me.
AIESEC Uganda
Led the chapter through 46% growth in active membership; managed budget and partnerships with Baraka Community Foundation and Happy Times Childcare Initiative.
Uganda Industrial Research Institute
Three months on hardware projects between research seasons — sensor integration, PCB design, and a working security-system prototype.
AIESEC International · Information Management
Resolved global platform issues, translated user feedback into product fixes, and helped close the loop between members and engineering.
Uganda Christian University
AI, data structures, big-data analytics, robotics, embedded systems. Graduating July 2026.
Short courses and certifications taken along the way
Amazon Web Services
IBM
Stanford Deep Data Research Center
ALX
Cisco
Side reading, running parallel to the work.
“Chance favors the connected mind.”
Lessons from AI about evolution, computing, and minds
“The best way to predict the future is to invent it. The second best is to talk with someone who is.”
— a working belief
Luganda — so little done, so much more to do.