Responsibilities include the development and maintenance of data science infrastructure, the exploration and implementation of technologies that enable big-data analytics, and the continual research and application of machine learning techniques on these data, particularly out-of-core methods. Technologies employed include ElasticSearch, Spark, MongoDB, LXD/Docker, HDF5, R/Python, Cassandra, Hadoop, HDFS, BigQuery.Additional responsibilities include cloud infrastructure management, instance security (typically Unix/Linux based instances), automated deployments and the scaling concerns around data throughput and data access.

Presenting masters and postgraduate diploma courses in big data technologies within the data science program offered by the Industrial Engineering Department at Stellenbosch University.Topics covered include; data and model life cycles; processes, teams and the big data technology landscape; cluster and distributed computing; RDBMS, NoSQL; data storage architectures (e.g. lakes and feature stores); big data processing (MR, Spark, Real-time with Beam); big data analytics, tools and principles; big data architectures.

This part-time endeavour allows us to provide value to researchers in academia by assisting in data aquisition, data storage, and data analysis in tasks that require specific technical knowledge (particularly for large datasets). For instance, one such project required us to build a number of tools/algorithms around Tesseract in order to improve optical character recognition on a large data set.

Responsibilities included the application of standard data analysis methodologies (data exploration, data preparation/preprocessing, model selection and model validation) in building models for customer acquisition. The resulting models were deployed within containerized micro-services that allowed for easy scaling. Successful modeling techniques employed included conditional random fields, support vector machines, neural networks and simple regression models.

This position required a division of my time between two subsidiaries of the aerospace group, namely Space Advisory Company (SAC) and Space Commercial Services Global Information (SCSGi). For SAC I contributed to the big-data storage solution for the Square Kilometre Array (SKA). We were tasked with determining if custom built storage pods would satisfy the SKA's use-cases (a high speed throughput solution and petabyte archive solution). This involved writing a test harness to measure data throughput for various disk configurations for high speed data throughput. The archive storage solution required a CEPH object store be prototyped. This work was done on pod hardware, as well as in a high performance computing cluster.For SCSGi I worked primarily on a natural language processor using conditional random fields. This work required the processing and cleansing of data from a mongoDB database, the establishment of a ground-truth dataset for modeling, and the selection of features and optimisation of the classifier.

Worked within an agile development team that worked exclusively on the company’s platformtechnology stack. Responsibilities included Java application server design, development, testing and deployment, after-hours operational support for both enterprise and platform technologies, and the design and implementation of reporting/analytics servers. The latter responsibility necessitated the design and analysis of data warehouses and data marts, as well as, the creation and maintenance of custom distributed extract, transform and load (ETL) processes for managing data extraction from multiple sources. Tools and technologies used throughout include: UNIX, Amazon EC2, Glassfish, J2EE, Maven, REST, MySQL, TestNG (unit, integration and functional testing), and Jenkins.

Oversaw the migration of most users to Linux and deployed various services (LDAP, CUPS, SAMBA, NFS and chat servers). Projects carried out during this time included: a client-server camera security system using old computers and USB webcams (written in C using OpenCV and ffmpeg libraries); the addition of a printer accounting component to the CUPS software using their API (written in C); and the design of a disk-less computer cluster and the procurement of the necessary components. The cluster utilised a TFTP and DHCP server that provided a minimal kernel to the disk-less nodes. Shared disk space was made available via NFS. Ganglia was used for cluster reporting and all remaining interactions were programmed in Bash.Furthermore, I assisted in the distribution and design of programs used in solving difficult combinatorial problems.

Developed a C ++ program for the express purpose of interfacing and analysing a state-purchased air-defence product.

Developed a non-uniform rational B-splines tool-kit (in Java) for the visualisation of dykes andfaults within existing seismic modelling software (Feb 2005 - Jul 2005). This included preliminary research into three-dimensional surface estimation from point clouds. A feasibility study of database technologies for use within the company (Aug 2005 - Dec 2005)

Work was performed during the university holidays within this period in a team selected from undergraduate computer science students. It comprised two projects. Firstly, the Ecosim Project involved the simulation and visualisation of a virtual world (Dec 2001 – July 2003, www.cs.sun.ac.za/~ecosim , programmed in Objective-C and Python) and secondly, the Ant Routing Simulation, applied the natural paradigm of foraging ants to find optimal routes in IP networks (Dec 2003 – July 2004, programmed in Java).

Duties included the day to day running of the hostel and organisation of events.

Responsible for processing cash receipts and cash payments for BDO branches.

Responsible for receiving clients, operating the switchboard, dealing with general enquiries, manag-ing mail and keeping project and personnel information up-to-date.