From capacity building to deployed solutions - the “end-to-end” Data Science Africa approach
Written by GeDIA member Ciira Maina // Program chair for Data Science Africa (DSA) and lecturer of Electrical Engineering at Dedan Kimathi University of Technology, in Kenya.
In recent years, advances in Artificial Intelligence (AI) have led to the development of technologies that were in the realm of fantasy just a decade before. We can now imagine driverless cars and interact with complex technologies through speech. AI has already had a great impact on the lives of people around the world. Unfortunately, AI applications in Africa are not as widespread, despite the huge potential for positive impact. For this to change, Africans must be at the forefront of AI initiatives on the continent for only the wearer of the shoe knows where it pinches — it takes a local to have the intimate knowledge required to develop local solutions.
A key ingredient in realising the ideal situation — where Africans take the lead in solving local problems — is the creation of a large pool of competent AI practitioners.
Recently, a number of initiatives have emerged that aim to address the AI skills gap on the continent. Since 2015, I have had the pleasure of being part of Data Science Africa (DSA) which has been at the forefront of increasing the data science and AI expertise in Africa.
DSA events include a summer school where participants are introduced to machine learning methods and data science, as well as a workshop where practitioners and researchers give presentations on their work applying AI to real world problems. Over the years we have had several presentations of interesting work being done in Africa in areas such as agriculture, healthcare, and environmental conservation.
At DSA, we realised very quickly that data would be at the heart of any progress made in applying AI to problems on the continent. This meant that we needed to introduce participants to data collection and creation of data sets usable by machine learning algorithms. A key driver of this was a partnership with ARM that began in 2017, which allowed us to obtain sensors we could deploy to obtain data. These data could then be used to introduce machine learning concepts. We call this the “end-to-end” data science approach.
In 2018, we held a 'train the trainer' event at Dedan Kimathi University of Technology (DeKUT) in Nyeri, Kenya. This event included a six day summer school and a two day workshop. During the summer school, theoretical sessions were accompanied by practical coding sessions, and deployment of sensors to collect data. The systems included a camera trap network deployed at the university conservancy to collect images of wild animals, an air quality monitoring sensor network, and a greenhouse monitoring system.
Following the 2018 'train the trainer' event, participants were encouraged to explore sensor deployments to solve problems in their local context. At DeKUT, a group of students deployed a greenhouse monitoring system and a river level monitoring system, as shown in the images below.
In June 2019, we moved things a notch higher. A special board which contained a number of sensors capable of measuring several parameters (including temperature, humidity, light intensity and barometric pressure) was developed. The board has connectors for wind speed and air quality measurement too. At the end of the event, these boards were distributed to participants from several African countries, with the aim of inspiring several local projects in collecting weather data with potential applications in agriculture.
At Deep Learning Indaba in Nairobi, I led a 'maker session' together with a number of my students. In the session, we introduced participants to hardware programming using Mbed, IoT connectivity using LoRa, time series databases, and machine learning methods for time series data.
The hardware used during the session included the STM32 Nucleo board and a custom LoRaWAN shield to transmit data from temperature and soil moisture sensors over LoRa to databases running on local machines. The code for this session is available on GitHub.
Capacity building is about empowering the next generation. In preparation for the Indaba session, four of my students Jared Makario, Yuri Njathi, Humphry Shikunzi, and Collins Emasi actively participated in developing and testing the material for the session. This involved testing the hardware and going through the code to ensure everything was clear.
During the session, they were able to help participants and ensure things went smoothly. As we move forward, we need to reach more and more young students to ensure that they obtain the necessary skills in AI, machine learning and data engineering. At DSA we are committed to this and we hope that the future will validate this approach.
In addition to empowering the next generation of data scientists, it is important to ensure diversity among participants and to ensure different learning styles are catered for.
To this end, DSA is keen to explore ways to increase the participation of women in future DSA events to help close the gender representation gap in data science education. The partnership between DSA and GeDIA in organising a gender equality-themed data science summer school and workshop for 2021, will go some way in achieving this, and ensuring that the approaches we take are appropriate.
[This text is adapted from an original blog post previously published on Ciira's website]