Omnien Platform
AI, collaboration, omnichannel, communication, virtual assistant, knowledge management
Omnien
Omnien is a state-of-the-art digital platform that enables effortless knowledge management, collaboration, and communication for businesses, communities, and individuals alike. Omnien gives you the flexibility to engage through your preferred channel, at your preferred time, and in your preferred way, whether it’s through private messaging, voice calls, or video calls. Omnien’s AI and automation are designed to quickly and accurately answer your questions, directing you towards the appropriate resource efficiently on your first attempt. Omnien’s collaborative tools and intuitive user experiences can assist you in enhancing your engagement and productivity.
Global Technology Outlook
Technology strategy
IBM
For the past 4 decades, IBM Research has created the Global Technology Outlook. The GTO is an annual opportunity to influence the direction of the company by conducting an in depth and broad analysis that identifies and evaluates technology inflection points, growth areas and business opportunities. The societal and business applications in which these technologies could be used, and their potential impact is considered. The theme for this GTO was “The Urgency of Science in a Post COVID-19 Crisis World”. This GTO describes the urgency of science along with the opportunities for accelerating and scaling discovery using the scientific method to produce a more agile and rigorous approach to tackle complex problems in science, business, and society.
Neuro-Symbolic AI
AI
IBM
Most of the neuro-symbolic reasoning research does simple proof of concept, working with small examples, strong language bias, predicate templating, and strong supervision. The primary theme of this project is to actively demonstrate the benefit of incorporating symbolism into connectionist solutions, and the main hypothesis is that the benefit principally derives from the transferability of reasoning mechanisms (e.g., analogical argumentation, entailment, and chained inference). New demonstrations of the superiority of neuro-symbolic learning and reasoning will spur development of complementary approaches that combine deep learning advancements with symbolic AI to express their strengths and supplement their weaknesses.
TB Genomics
AI, healthcare
IBM
In 2016, an estimated 600 000 cases of multi-drug resistant (MDR)/rifampicin resistant (RR)-TB and over 8 000 cases of extensively drug resistant (XDR)-TB were recorded globally. The solution was to develop deep learning models that identify novel genetic mutations conferring resistance to TB medication and compensatory mutations. Developing a classification system that can distinguish between drug sensitive (DS), multidrug (MDR) and extensively drug (XDR) resistant TB.
Automatic machine learning based classification of cancer pathology reports
AI, healthcare, NLP
IBM
The main objective of the research was to develop machine learning methods that can automate processes used by cancer registries including report labelling using the globally recognized International Classification of Diseases for Oncology (ICD-O), 3rd edition. Additionally, natural language understanding techniques are used to mine and extract the important medical concepts from the pathology reports.
Time and motion of TB and HIV clinic services in Johannesburg, South Africa
AI, healthcare, IoT
IBM
Existing methodologies for performing time and motion studies rely on subjective manual clinical observations by healthcare workers, are subject to observation errors and are not based on objective measures. The automated time and motion deployment is composed of; a set of wearable electronic devices, which patients and healthcare workers will wear, fixed location electronic readers to collect data on waiting time for services, a mobile application for downloading data from the wearable devices and a Cloudant database to store the data.
Cognitive-Assisted Interactive Labelling
AI, active learning
IBM
There is a consistent demand from data scientists and machine learning practitioners for a production-quality annotation tool for image datasets. The goal was to develop the best cognitive-assisted interactive labelling pipeline by using the best-performing methods in unsupervised feature extraction, transfer learning and dimensionality reduction.
RFI Excision and Anomaly Detection
AI, astronomy
IBM
Select well-defined narrow problems in radio astronomy with existing data resources available and to which a firm hypothesis for use of machine learning exists.
- Radio frequency interference (RFI) masking: Utilize deep learning techniques of pooling for fast sensitive RFI detection.
- Bayesian radio interferometry: Conducted computational research for radio telescope inclusion analysis for very large baseline arrays.
- RFI Simulation: Enable the generation of groundtruth RFI for the MeerKAT instruments by using Generative Adversarial Networks to learn from sparse data.
A system for accelerated language learning in classrooms
AI, education, NLP
IBM
To fix the education crisis, we proposed that we use daily reading and written exercises in classrooms to improve English proficiency. Fix the English, and we fix the language barrier. Fix the language barrier, and we have a shot at fixing the education crisis. To facilitate daily written exercises, we implemented a peer assessment platform that allows students to provide feedback to one another on their work, assisted by teachers and Artificial Intelligence tools.
Air Quality Evaluation and Forecasting
AI, environment, sustainability, IoT
IBM
With air quality evaluation the objective was to unveil how well intervention strategies to improve air quality were working. This involved looking at historical and live data from air quality monitoring stations and comparing the data to air quality standards targets. This was fundamental to assessing the level of health risk associated with local air quality.
For air quality forecasting the objective was to perform air quality forecasting with high accuracy and at a high resolution. This would provide a valuable tool for planning and decision support and would encourage a pro-active approach to dealing with potential incidents where air quality will be at unfavourable levels.
‘Sensing and Intelligence’ track at IEEE AFRICON 2017
Track chair, peer review
IEEE
I was a track chair for the ‘Sensing and Intelligence’ track at IEEE AFRICON 2017. AFRICON is the IEEE’s flagship conference in Africa. I also organised IBM’s participation as a platinum sponsor of the conference and a keynote speaker.
‘Applications of Machine Learning in Africa’ Workshop
Workshop Organiser
IEEE
I organised a workshop titled ‘Applications of Machine Learning in Africa’ that was held as part of the proceedings of the IEEE AFRICON 2017 conference
Women Pushing Boundaries in Science and Technology’
Organiser
IEEE, IBM
I conceived the idea and coordinated the organisation of a conference titled ‘Women Pushing Boundaries in Science and Technology’ that was co-organised with the South Africa chapter of the IEEE Women in Engineering affinity group. I arranged for many prominent speakers to participate including the CEO of the IEEE.
Expert Technical Review
Peer review
Various
I have been an expert reviewer and served on the technical program committee for several conferences and journals locally and internationally. These include the International Conference on Adaptive Science and Technology (ICAST, 2009), the Southern African Telecommunication Networks and Applications Conference (SATNAC, 2011, 2012), the IEEE Transactions on Industrial Informatics (TII, 2013, 2014), IEEE Transactions on Industrial Electronics (TIE, 2013), and the Springer Journal on Wireless Networks (WINET, 2013).
Solar Photovoltaic Component Estimator
Mobile application development, web API
IBM
The goal of this project was to design and build a consumer-friendly app that empowers homeowners to design their own battery-based PV system. The app selects components from a live catalogue, which is based on the availability of equipment from nearby retailers and optimises the system on price and performance. The app is accessible via a web app which can be viewed on a mobile, tablet, or desktop device.
Asset Maintenance Optimisation in Electricity Distribution Networks
Optimisation, asset maintenance
IBM
The objectives of this project were to optimise the management of inventory in electricity distribution networks such that there can be greater operational efficiency, reduced unplanned outages and lower operational costs. The approach planned was perform concept work for the design and development of an integrated system for optimally managing asset inventory including a model and application that does away with cases of excess and inadequate inventory and creates a distinction between how critical and non-critical stock is handled.
Data Driven Expertise Matching
Machine learning, dynamic teaming, graph theory
IBM
The goal was to create a system that continually outputs a representation of all the skills, competencies, and disciplines that best describe the work within the company. Such a system would be used to help shape company strategy, foster dynamic teaming, and to inform hiring and professional development.
Robust 2d/3d Sonar Imaging System for Harbour Surveillance
Sonar imaging, 2d/3d visualisation, human machine interface, system integration
CSIR
I worked on the system integration, configuration, and human machine interface of a robust sonar underwater imaging system. This involved upgrading legacy software and coordinating development efforts with other team members. There were tight deadlines in this project, but all my deliverables were met on time and within brief. Different application profiles could be created and managed using the software between different sessions. Seamless changeover between 2d and 3d visualisation modes was achieved as well as between different signal processing algorithms. Past application sessions application profiles could be saved and later opened and 3d scenes could be stored for later analysis. Technologies used included C++, OpenSceneGraph, OpenGL and QT.
Visual Surveying Platform
Machine learning, data capture, analysis, and visualisation
CSIR
I wrote a software tool to capture data from Gig-E vision cameras and GPS units connected to a custom sensor rack. I assisted in developing a software tool to generate a dataset of images that can be used to train an Artificial Neural Network (ANN). I wrote an implementation of a feed-forward ANN for the purpose of classifying road surface distresses. As part of the project there was a desktop application which I developed that provided an interface with which a user could view the location of detected road surface distresses on a map and create job cards for them.
Water Quality Monitoring System
Service oriented architecture, real time monitoring and control
CSIR
I wrote a software suite for a water quality monitoring system. One component of the software suite was a server to communicate with remote water quality monitoring stations, logging readings into a database that I setup and coordinating firmware upgrades of the units. There was also a web-application written using the Google Web Toolkit with which users could monitor remote stations in real time. Another component was a configuration utility that interfaced with the database to setup new remote stations, their capabilities, and what actions they must perform. For easy monitoring of stations in the field, I also wrote an Android mobile application. A service-oriented architecture was used, and a REST web service was written.
Mining Safety Platform
Mobile application, teleoperation, telemonitoring
CSIR
The monitoring and control application for the Underground Autonomous Mining Safety Platform (UAMSP) is a mobile application that runs on a tablet computer with Android Honeycomb (3.0) or higher. The code for the application was written in Java and resource files for its user interface and application constants were written in Extensible Mark-up Language (xml). The software allows a user to monitor and control the UAMSP. It provides functionality to steer the UAMSP, move it backwards and forwards, adjust its speed, lift, or lower its flipper arms, as well as pan and tilt its on-board camera. It is possible to view a live video stream from the UAMSP as well as view reports on the status of its batteries, motor controllers, inertial measurement units and environment sensors. I was lead developer on this project.
Automated Blood Diagnostics System
Cloud service
CSIR
I developed a server and cloud service to assist in the analysis and monitoring of samples taken by a medical diagnostic device. I also developed a monitoring tool to visual how the diagnosis is progressing on the server. Technologies used included WCF, WPF, C#, MySQL.
Grader Training Simulator
Grader, virtual reality, simulator
5DT
I led the development of a Caterpillar CAT140M grader simulation module for use in the flagship road construction simulator. This included computer graphics, dynamics, kinetics, kinematics, and modelling. I simulated things such as ground interaction, engine power and torque, fuel consumption, gear ratios, driving force, suspension, self-aligning torque, drag resistance, rolling resistance, braking force, collision impulses, animation, kinematics, vehicle gauges, vehicle controls, sounds and particle systems. I also worked on in program operator assistance aids for use in training in the form of 3D visualisations. I developed a way of handling ‘passive dynamic entities’ which are objects in the virtual world that the vehicle can interact with. This simulator has been used on sites in countries such as Ethiopia and Australia
Vehicle Operator Computer Based Training (CBT) Course
Computer based training
5DT
I was responsible for the generation of course content and a syllabus for a haul truck computer-based training program. Such a course was targeted at potential operators of haul trucks and covered topics such as where and how a haul truck is used, how it compares to other types of trucks; the components of a haul truck including its instruments and controls and how they work; safety aspects of operating a haul truck; rules of the road when operating the truck in a surface mining or quarry environment; procedures and processes of using a haul truck in a production environment including topics such as pre and post shift inspections; loading, hauling, queuing and dumping; what to do when things go wrong; and finally advanced haul truck topics such as optimal conditions for haul trucks.
Roof Bolter Training Simulator
Roof bolter, virtual reality, simulator
5DT
I assisted in the development of a simulation module for a Sandvik DS201L-M underground drill rig which was used in the flagship underground mining simulator by writing code to simulate the vehicle physics. This included computer graphics, dynamics, kinetics, kinematics, and modelling. This simulator was initially used at a platinum mine in Twickenham.
Haul Truck Training Simulator
Haul truck, virtual reality, simulator
5DT
I developed several haul truck simulation modules for use in the flagship mining simulator as part of a team which I led for some of the projects. These included the CAT793D (2009), CAT777F (2009) and CAT777D (2010) Caterpillar haul trucks from as well as the Fermel NL0068 (2008) articulated dump truck. This included computer graphics, dynamics, kinetics, kinematics, and modelling. I simulated things such as engine power and torque, fuel consumption, gear ratios, driving force, suspension, self-aligning torque, drag resistance, rolling resistance, braking force, collision impulses, animation, kinematics, vehicle gauges, vehicle controls, sounds and particle systems. I also worked on in program operator assistance aids for use in training in the form of 3D visualisations. These simulators have been used on sites in various countries worldwide including Australia, Brazil, South Africa, and the U.S.A.
Push2Talk Mobile Phone Application
Adaptive multi-rate codec, push2talk
LUUK
I developed a Push2Talk walkie-talkie type voice chat application that used the phones data connection. The software was targeted at cell phones that ran the Symbian operating system (OS) which was supported by several mobile phone manufacturers such as Nokia, Sony-Ericsson, and Siemens. The VoIP solution allowed users to carry out voice conversations using the data network instead of the traditional voice carrier of the mobile communication service provider. The software had additional features to enable voice mail storage and retrieval, caller identification and contact list creation. Communication was on a half-duplex basing hence the push to talk operation. Voice data was encoded using the adaptive multi-rate (AMR) codec to allow for voice compression and reduce the data payload. Technologies used included C++, Carbide IDE, and the Symbian S60 framework.
Web Based Fleet Tracking System
Reverse geocoding, reverse engineering
LUUK
I worked on the communication back-end of a web-based fleet management system. This included deciphering the communication protocol of a vehicle monitoring unit which had very little documentation which was primarily in Russian. The unit that recorded a vehicle’s GPS coordinates other pertinent details on the state of the vehicle (acceleration, braking, door open etc.) and sent the data in binary form using the GSM network which the software had to read. I also wrote software to read shape file information for use in online maps and reverse geo-coding.
Building Automation and Management using Wireless Sensor Nodes
Wireless sensor networks
University of Pretoria
A building automation and management system that makes use of wireless sensor nodes was developed by me. The system had various hardware and software components that were used to realize its functionality. A custom plug-in sensor module was developed as well as a custom central actuating unit. MicaZ wireless motes were used. The environment could be remotely monitored and controlled via a gateway using a PDA as well as a web interface which also provided a video feed. Energy usage statistics were available and various actuation rules could be set. Technologies used included Eagle, NesC, C#, Java, AJAX, Java SDK, TinyOS, JMF and the .Net Compact framework.
Teaching and Research
Lecturing, tutoring, mentoring, researching
University of Pretoria
I was involved in teaching and research at the University of Pretoria in several roles. I have assisted in teaching in several modules namely Physics, Design Patterns in C++, Computer Networks, Design and Manufacturing, and the Final Year Project. I have conducted research as part of the Advanced Sensor Networks Group. I was affiliated with the Centre for Telecommunication Engineering for the Information Society (CeTEIS). I have served as a judge at the final year engineering student competition and as an external examiner for some undergraduate and postgraduate modules.