The 2026 MIT808 Exhibition Is Here — 42 Students, 21 Projects, One Common Thread

Our Masters in Big Data Science students tackled climate policy, cancer genomics, parliamentary AI, and African language models. Here’s what they built.

Every year the MIT808 Data Science Capstone module asks a deceptively simple question: can you take everything you’ve learned, find a partner with a real problem, and build something that matters?

This year, 42 students said yes. The result is the 2026 MIT808 Capstone Exhibition, now live at https://up-mitc-ds.github.io/808exhibition.

Below is a tour of what they built.

🌍 African Climate NLP

Five teams dove into the complex world of climate governance documents — UNFCCC national submissions, South African Hansard records, and SADC policy frameworks. Their work is a reminder that the hard part of climate action isn’t just the science; it’s the language.

Projects examined thematic classification across policy corpora, bias in LLM-generated climate recommendations, multilingual asymmetries in how different countries frame their commitments, and colonial framing embedded in the language of climate governance itself.

🛩️ UAV & Population Estimation

How do you count people in an informal settlement when a census is too slow and too expensive? Three teams answered this question using drone imagery of the Melusi settlement in Atteridgeville, combining deep learning segmentation models (U-Net, DeepLabV3) with Bayesian and Gaussian Process regression to estimate population from above.

The implications for planning, resource allocation, and disaster response in underserved communities are significant.

💨 Air Quality Forecasting

Ground-level ozone on the South African Highveld is a serious public health concern — and three teams built tools to predict it. Their pipelines range from same-day alert systems using Random Forest and XGBoost, to 3-hour lead-time forecasts for Secunda, to 24-hour Highveld-wide predictions delivered through Streamlit dashboards designed for real operational use.

🏛️ Parliamentary Intelligence

South Africa’s parliamentary record is vast, complex, and consequential — and three teams built tools to make it navigable for investigative journalists and civil society. Their systems combine topic clustering, intent classification, abstractive summarisation, and RAG-assisted search across PMG data. The goal: make accountability journalism faster.

⚖️ African NLP Governance

The African NLP ecosystem is growing fast — but who owns the data that trains its models? Three teams audited 249 African NLP datasets for copyright compliance and licensing risk, using rule-based scoring, ML classifiers, and unsupervised clustering. The findings reveal systemic governance gaps, and their Streamlit dashboards are designed for real-time risk auditing by dataset curators.

🔬 Prostate Cancer Decision Support

The South African Prostate Cancer Study (SAPCS) is generating rich whole-genome sequencing data — and three teams built the clinical dashboards to make sense of it. Projects cover risk stratification, molecular driver visualisation, and genomic integrity profiling, aimed at supporting clinical decision-making in South African oncology.

🗣️ African Language AI

One team tackled a quietly urgent problem: late-stage rabies reporting in communities where health education isn’t available in their home language. Their project evaluates open-access LLMs for Sepedi translation, laying groundwork for a rabies awareness chatbot that could reach communities often left behind by health systems designed around English.

A Note on the Course

MIT808 is taught as part of the Masters in IT (Big Data Science) programme at the University of Pretoria. Students spend the year working with real partners on real problems — the module is less about coursework and more about what it actually feels like to be a data scientist.

The 2026 course was led by Dr Olaperi Okuboyejo and co-coordinated by Dr Abiodun Modupe. More information at dsfsi.co.za/mit808.

Browse the full exhibition — including posters, videos, and an interactive project map — at: 👉 https://up-mitc-ds.github.io/808exhibition

Questions or interest in partnering? Reach us at dsfsi.info@up.ac.za.

Stay connected with our work:

• DSFSI: https://linktr.ee/dsfsi
• AfriDSAI: https://linktr.ee/afridsai