Speakers - #indabaX

Ahmed Mohammed Ahmed

Grokking in neural networks and the role of regularisation

Day 2 - 15h00 - 16h30

Lecture

Alaa Aldein Mohamed Suliman Ibrahim

University of Kwazulu-Natal

Transfer Learning-Based Method with Extreme Gradient Boost Classifier for Surface Defect Classification of Hot-Rolled Steel Strips

Day 2 - 10h30 - 12h00

Lecture

Amy Bray

Zindi

Competitive Machine Learning to Improving your CV

Lecture

Avi Moodley

Council for Scientific and Industrial Research (CSIR)

NLP Solutions for Practical Industry Use Cases

Day 4 - 13h00 - 14h00

Lecture

Besman Guesmi

DETSlideNet: Dual-Encoder SwinTransformer for Scalable Landslides Detection and Monitoring

Lecture

Bigomokero Antoine Bagula

University of the Western Cape (UWC)

Artificial Learning: From Federated Learning to Diversity Learning

Day 2 - 15h00 - 16h30

Lecture

Talk description

As Artificial Intelligence continues to progress, the underlying architectures that support learning across distributed, privacy-sensitive environments must also evolve. “Federated Learning (FL)” has emerged as a powerful framework for privacy-preserving, decentralized model training. However, its conventional implementation assumes homogeneous data distributions and identical model architectures across all clients—assumptions that rarely hold in real-world scenarios. This talk introduces the concept of “Diversity Learning” within federated architectures. In this approach, clients are grouped into clusters based on the nature of their data and operational roles. Each cluster is trained using a model architecture best suited to its specific context—whether temporal, spatial, or multi-modal—giving rise to the “Federated Diversity Learning (FedDivL)” paradigm. FedDivL embraces both data and model heterogeneity, enabling more flexible and context-aware learning. However, this diversity introduces new challenges in aggregating knowledge across heterogeneous clusters. To overcome these, FedDivL requires the development of “adaptive fusion strategies” that move beyond conventional averaging techniques. These strategies enable more informed and reliable global decision-making. Selected fusion strategies will be presented in this talk. Through use cases in network intrusion detection (e.g., SDN-based DDoS attacks) and smart infrastructure systems, the talk will demonstrate how diversity-aware federated learning enhances system robustness, adaptability, and generalization—particularly in dynamic and resource-constrained environments. Participants will gain insight into the limitations of traditional FL and explore how embracing diversity in both data and model design can unlock the next generation of collaborative, intelligent systems.

Speaker bio

Prof. Bigomokero Antoine Bagula is a professor of Computer Science at the University of the Western Cape (UWC), South Africa, where he leads the Intelligent Systems and Advanced Telecommunications (ISAT) Lab. He also holds a professorial role at Université Nouveaux Horizons (UNH) in the DRC. He holds a Ph.D. in Communication Systems from KTH (Sweden), and MSc degrees from UCL (Belgium) and Stellenbosch University (South Africa). His research spans Artificial Intelligence, Federated Learning, Big Data, Cloud/Fog Computing, High-Performance Computing, IoT, and Network Softwarization (NFV/SDN). With a strong focus on applying machine learning to distributed and resource-constrained environments, he contributes to advancing digital infrastructure in the Global South.

Charl Botha

AI Safety

AI Progress: Promise or Peril

Lecture

Craig Parker

Wits Planetary Health Research

Beyond Weather Warnings: Machine Learning Reveals the Hidden Biology of Climate Health Vulnerability

Day 4 - 14h30 - 15h30

Lecture

David de Villiers

MWR CyberSec

Overlooked and Undersecured: Prompt Injection Attacks on LLM Applications

Day 2 - 10h30 - 12h00

Lecture

Dr Sebonkile Thaba

University of Johannesburg

From Supply Chains to Smart Solutions: Empowering Non-Technical Students with Blockchain Innovation

Day 2 - 10h30 - 12h00

Lecture

Dr. Axel Tidemann

Takealot

Machine learning at Takealot

Day 3 - 09h00 - 10h00

Lecture

Dr. Gavin Rens

Stellenbosch University

Symbolic Methods in Artificial Intelligence and Machine Learning

Day 3 - 10h30 - 12h00

Fundamentals

Dr. Hossein Masoumi-Karakani

University of Pretoria

Data Culture Rising: Literacy for the Next Generation

Day 4 - 10h30 - 12h00

Fundamentals

Dr. Joan Byamugisha

IBM Research - Africa

Evolution of NLP Methods, and their Fit-for-purpose

Day 2 - 15h00 - 16h30

Fundamentals

Dr. Karen Walstra

AI School Policy: An Urgent Need

Day 4 - 14h30 - 15h30

Lecture

Dr. Lili Rademan

Centre for Research on Evaluation, Science and Technology (CREST), Stellenbosch University

Beyond the Poster: Practical Presentation Skills for Researchers

Day 1 - 09h00 - 11h00

Workshop

Dr. Sara Andreotti

Stellenbosch University / SharkSafe Barrier Pty / SharkWise Pty / SCAR Trust

Workshop: Building Science Communication Capacity in Africa

Workshop

Dr. Werner Janse van Rensburg

Centre for High Performance Computing

Introduction to NICIS: Your Compute (CHPC), Data (DIRISA) and Network (SANReN) Infrastructure Services Partner

Day 4 - 13h00 - 14h00

Lecture

Dylan Geldenhuys

Sand Technologies

Building a Digital Twin for Telco: Optimizing Customer Satisfaction with Causal ML - Simulating interventions, predicting customer impact, and guiding network decisions at scale

Day 4 - 14h30 - 15h30

Lecture

Francois Meyer

University of Cape Town

Improving Text Generation for Low-Resource South African Languages

Day 2 - 15h00 - 16h30

Lecture

Francois Naude

Institute of Mine Seismology

Hackathon / Augmentation and Training Strategies for Improved Phase Arrival Picking in Microseismic Monitoring

Day 2 - 15h00 - 16h30

Lecture

Gabriela Penelopé Carolus

Division of Health Systems and Public Health, Department of Global Health, Stellenbosch University

Ground Truth: Blending Clinical Evidence and Unstructured Deep Learning for a Culturally Responsive AI Health Innovation

Day 4 - 14h30 - 15h30

Lecture

Talk description

This presentation introduces StoryRx, an innovative, AI-driven digital storytelling health communication tool designed to support Type II diabetes management in diverse South African contexts. Due to the historical underrepresentation of African languages, cultures, and environments in generative AI (GenAI) training, many existing models struggle to effectively address content relevant to Africa. StoryRx employs advanced training and prompting strategies tailored to the unique linguistic and cultural landscape of the Western Cape, thereby addressing significant gaps in current AI capabilities. We will explore both the limitations and opportunities within GenAI, highlighting ongoing initiatives in the African ecosystem, such as InkubaLM, UlizaLlama, and AfriCOMET, while also acknowledging persisting challenges in image and speech models. Through our collaborative project, which leverages advanced retrieval-augmented generation (RAG) pipelines alongside structured clinical evidence and culturally nuanced narratives, we demonstrate improved evaluation methods, including human-centered and expert-in-the-loop approaches. Our findings highlight the importance of critical cultural and contextual considerations for creating AI-driven stories that resonate with their intended audiences. Ultimately, StoryRx exemplifies how integrating structured, evidence-based medical guidelines with culturally responsive, unstructured narrative data can lead to ethically sound, clinically accurate, and contextually relevant AI health applications, paving the way for equitable and scalable GenAI solutions worldwide.

Speaker bio

I am a Research Fellow and Research Project Manager at Stellenbosch University, specialising in the intersection of artificial intelligence, software architecture, design, public health innovation, and human-computer interaction. My research focuses on the co-design process for culturally sensitive, AI-driven digital storytelling health interventions and developing digital health solutions to support Type 2 diabetes management in low—and middle-income African contexts. I am cultivating dual expertise in Health Data Analytics, utilising Python, R, and machine learning methods, alongside Software Engineering and AI Architecture. I also pay particular attention to the ethical dimensions of digital health systems in resource-constrained environments. My work is dedicated to designing, evaluating, and scaling digital health solutions through participatory approaches and digital randomised controlled trials (RCTS). In the short term, I aim to publish in peer-reviewed journals, present at international conferences, and prepare for early-career grants and postdoctoral fellowships. Ultimately, I aspire to lead initiatives within organisations such as UNESCO, WHO, UNU, or the CDC, advancing digital public health innovation across Africa.

Jason Max Chalom

Building out a sustainable and production ready AI software solution

Day 3 - 10h30 - 12h00

Lecture

Jeroen Van Goey

InstaDeep

Machine Learning Interatomic Potentials (MLIP)

Day 3 - 10h30 - 12h00

Lecture

John Matogo

IBM

Skills Build Resources

Day 3 - 13h30 - 14h30

Lecture

Luke Taylor

DeepGate

Why AI is Power-Hungry - Lessons from the Brain and How We Can Fix It

Day 3 - 10h30 - 12h00

Lecture

Mbalenhle Ntuli

Stellenbosch University

Workshop: Building Science Communication Capacity in Africa

Day 1 - 09h00 - 11h00

Workshop

Mr. Boyd Robert Kane

AI Safety

Employment after AGI: Paths, Dangers, Strategies

Lecture

Mr. Luther Chipembere

Stellenbosch University

Workshop: Building Science Communication Capacity in Africa

Day 1 - 09h00 - 11h00

Workshop

Mr. Naweed Khan

IBM Research Africa

The state of AI

Day 3 - 10h30 - 12h00

Lecture

Mr. Ndivhuwo Nyase

IBM Research Africa

Quantum Machine Learning 4 Africa

Workshop

Mr. Thapelo Maupa

Independent Researcher.

A friendly Introduction to the Mathematical Basics for Machine Learning.

Day 1 - 09h00 - 11h00

Workshop

Ms Katherine Rabik

Beyond the Poster: Practical Presentation Skills for Researchers

Day 1 - 09h00 - 11h00

Workshop

Mzwandile Mhlongo

Standard Bank

Building AI Agentic Workflows from Scratch: A Hands-On Tutorial on RAG, Text2SQL, & Tools

Workshop

Ntshuxeko Confidence Mathebula

Opti-Num Solutions

Intelligent Systems Engineering: Bridging AI and Simulation

Lecture

Prof Marlene Holmner

University of Pretoria

Ethics in the Age of AI: Balancing Privacy, Accuracy, and Access

Day 2 - 13h30 - 14h30

Keynote

Talk description

Title: Ethics in the Age of AI: Balancing Privacy, Accuracy, and Access Speaker: Prof. Marlene Holmner, Department of Information Science, University of Pretoria

As Artificial Intelligence becomes increasingly embedded in everyday life—shaping decisions, driving innovation, and redefining information systems—it also brings forth complex ethical dilemmas. This keynote explores the critical ethical dimensions of AI with a specific focus on privacy, accuracy, intellectual property, access, and security in the context of generative AI tools and large language models.

Drawing from real-world examples, pending global lawsuits, and emerging African AI policy frameworks, Prof. Holmner unpacks the dangers of data misuse, algorithmic bias, and information inequality. She highlights how AI systems, while powerful, can replicate societal prejudices, leak sensitive data, and exacerbate digital divides.

Through the lens of ubuntu and African values of fairness and transparency, this talk calls for a responsible, context-sensitive approach to AI development and usage. Attendees will walk away with a deeper understanding of ethical risks, practical recommendations for researchers, and the urgent need for policies that protect both individual rights and collective societal good in the AI era.

Speaker bio

Marlene Holmner holds a DPhil in Information Science from the University of Pretoria, South Africa, and is an Associate Professor and the current Head of the Department of the Department of Information Science at the University of Pretoria. Her research interests lie in the field of Information Ethics, Indigenous Knowledge, ICT for Development, Information and Knowledge Society, Institutional Repositories, Digitization, Mobile Technologies and Curriculum Development. Her work has been published in various academic journals including the South African Journal of Libraries and Information Science, Information Development, Innovation and Mousaion. Furthermore, she has published in internationally accredited journals such as Library Hi-Tech, Library Philosophy and Practice and Information Technology for Development, to mention a few. She has presented various papers and led panel discussions at national and international conferences such as ProLISSA, SCECSAL, IFLA, IATUL, ALA, iConference, ASIST and pre-ECIS and AMCIS SIG GlobDev workshops. She has also been the programme chair and co-chair at SIG GlobDev workshops and mini-tracks. She teaches courses related but not limited to these fields at undergraduate and postgraduate level and as the study leader and supervisor of several Masters and PhD students. Marlene is a member of various national and international organisations including, the Library and Information Association of South Africa (LIASA), the African Centre of Excellence for Information Ethics (ACEIE), the Association of Information Systems (AIS), the Association for Information Science and Technology (ASIST), and iSchool member to mention a few.

Prof. Francesco Petruccione

National Institute for Theoretical and Computational Sciences (NITheCS), Stellenbosch University

Quantum Foundations: From Qubits to Intelligent Algorithms

Day 4 - 09h00 - 10h00

Keynote

Prof. Mandla Gwetu

Stellenbosch University

Fundamentals of ML

Day 2 - 10h30 - 12h00

Fundamentals

Prof. Pius Owolawi

Tshwane University of Technology

Edge AI for Agriculture: Towards a Cognitive Farming

Day 2 - 09h00 - 10h00

Keynote

Ruan de Kock

InstaDeep

A friendly introduction to the Mathematical Basics for Machine Learning.

Day 1 - 09h00 - 11h00

Workshop

Saheed Jimoh

University of Kwazulu Natal

Evaluation of Wind Speed Regime Variability and Model Fit Performance Using Multi-Component Weibull Mixture Functions for Energy Potential Estimation

Day 2 - 15h00 - 16h30

Lecture

Samuel Oladejo

University of Johannesburg / Xero

ForecastRL: Neural Time Series Forecasting Integration with Risk-Aware Deep Reinforcement Learning for Financial Trading

Day 3 - 10h30 - 12h00

Lecture

Shane Josias

Stellenbosch University

Multimodal base distributions in conditional flow matching generative models

Day 2 - 10h30 - 12h00

Lecture

Stefan Dominicus

Takealot

Recommender Systems at Takealot.com

Day 3 - 10h30 - 12h00

Lecture

Stefan Strydom

Pocket Polyglot Mzansi: Machine Translation for Local Languages on the Edge

Day 2 - 15h00 - 16h30

Lecture

Stephan Middleton

Von Seidels

The Patent Landscape for Inventions relating to Artificial Intelligence

Day 2 - 10h30 - 12h00

Lecture

Stuart Reid

NOSIBLE

Indexing the whole web

Day 1 - 13h00 - 14h00

Keynote

Tegan Green

UX designer exploring the implications of AI for human well-being

Lecture

Thato Angelo Serage

Forecasting South African consumer price index using statistical learning models

Day 2 - 15h00 - 16h30

Lecture

Tony Onazi Oche

University of Pretoria/ AI Safety South Africa

Real-Life AI Incidents as a Testing Tool for AI Safety Policies in South Africa

Lecture

Wiida Fourie-Basson

Stellenbosch University

Workshop: Building Science Communication Capacity in Africa

Day 1 - 09h00 - 11h00

Workshop

Zara Schroeder

Research ICT Africa

Unconsented: The Gendered Cost of Generative AI in the Digital Age

Day 2 - 10h30 - 12h00

Lecture