By the time I arrived at Accra’s Kotoka International Airport it was just before midnight. I was two hours behind schedule and the sky was clear and studded with stars. Despite being exhausted from the seven-hour flight from London, I was buzzing with ... nerves or excitement? It was hard to say.
I was in Ghana for Data Science Africa (DSA). DSA is an annual event that is run by the eponymous non-profit, knowledge-sharing, professional group. It brings together leading researchers and practitioners working on data science and machine learning (ML) in or for the benefit of Africa.
The events provide training on state-of-the-art data science methods to students and others interested in developing practical skills to tackle local problems – often those affected by the very problems they’re trying to solve.
This year’s theme was end-to-end data science – otherwise known as a single-team approach to identifying a problem, gathering and analyzing the data, and determining the solution. Sometimes, the teams also build the hardware to gather the data: it's an impressive amount of work from such small teams.
My role was to deliver a machine learning workshop at Ashesi University the next morning – a morning that was getting ever-closer as I inched my way through customs lugging a full box of dev boards (generously donated by Sparkfun for the occasion). After failing in my mission to get a rental car, I grabbed a hotel room, flopped into bed and prayed that I’d have time to get to my destination first thing in the morning.
The day dawned earlier that I might have liked, but I grabbed some breakfast and gathered my teammates before making a dash for Hertz … and was rewarded with Marla the Mazda, the cleanest, shiniest car in town.
For the next 90 minutes, we traveled on dirt roads, thick with mud since being drenched by the last weeks’ rains. Marla had lost her shine en route, but she did us proud: we were 30 minutes late for the day, but we were in one piece—no accidents, breakdowns or unanticipated outages along the way.
On the road to DSA
During the journey, Cirra Maina, Arm Innovator and one of the original organizers, explained that the event would be a five-day, intensive deep dive on data science and machine learning (ML), with in-depth presentations of real projects affecting real lives. It was to be attended by around 100 students from across Africa.
Cirra is a significant believer in local solutions to global problems. As he says, “the foot knows where the shoe pinches” – that is, those living with the problems are the bet that is placed to solve them. It is this notion that forms the ethos of DSA.
The first three days were summer school sessions focused on undergrad students learning engineering or data science. Instructors and practitioners from Arm, Google AI, Facebook, the UN, and Turing Institute – among others – were all in attendance to share their knowledge.
The aim of the summer schools is simply to make it possible for students to move forward in any direction they choose, so everything from satellite imagery, game theory, reinforcement learning, and IoT is covered. These teachings plant the seeds that can grow into full projects, which then form the basis of next year’s presentations.
ML on the edge is a relatively new topic – especially in Africa – and I was looking forward to giving students an insight into the latest technologies, hearing about the challenges they face and thinking about how we could begin to tackle them. The breadth of information presented by the various instructors was vast; sessions ranged from introductions to Python and Jupyter notebooks to an electrical grid mapping tutorial.
One of the newest, most compelling developments that was discussed was TinyML, a community of engineers focused on how best to implement ML in ultra-low power systems – something of obvious advantage in places such as Africa where connectivity and power are limited and constrained environments are the norm.
Students hard at work
At last, it was my turn to step up to the front, to deliver The Workshop: a lesson in how to create a simple wake-word ML-classifier running on Arm Cortex-M that is based on the Sparkfun Edge Board.
(If you'd like to have a go yourself, you can find the instructions on Codelabs: AI on a microcontroller with TensorFlow Lite and SparkFun Edge.)
The boards were developed to allow ML to be run on the edge – that is, directly on the device, without need for a network connection. Being able to make decisions at the point where it is needed and not needing to rely on power and connectivity can make the difference between actionable ML and disconnected IoT.
In the real world, this could be used anywhere that instant decision-making is required, such as predictive maintenance. Arm AI Partner Shoreline IoT, for example, builds IoT devices that can detect damage or material fatigue directly on the device, alerting operators before problems occur. Since they're not wasting power by continually uploading data to the cloud, these devices can run for years on a single battery.
Running the SparkFun Edge board
The last two days consisted of presentations by the practitioners on data science work in Africa – typically projects that are implemented locally, with mentorship from other teams, researchers, or companies.
There were a couple of projects that particularly inspired me:
The benefits of these solutions are tangible and can make a real difference to people’s health and livelihoods. They really brought home to me the relevance of DSA to the United Nations’ Global Goals for Sustainable Development – the 17 goals agreed to by world leaders in 2015 that aim to create a better world for everyone by 2030. Arm is committing to supporting the Global Goals through our sustainability mission.
The United Nations' Development Programme's Katie Bernhard goes over the Sustainable Development Goals
If I needed further evidence, the UN’s Development Programme’s sessions on their initiatives, such as Pulse Labs – which aim to pioneer new ways to use Big Data to support development goals – and the new UNDP Accelerator Labs – which look to develop radical new solutions to tackle the complexity of current development challenges – show how analyzing challenges locally, and working together to devise scalable, tech-based solutions, can help us solve them globally.
It was then that I had a revelatory moment: DSA Africa isn't about presenting novel approaches for novelty’s sake, or ‘toy’ problems being worked on by a company to showcase some advancement in theory. It's about real people and real communities using applied ML and data science to tackle problems that directly affect them. Here, data science and ML have the power to change lives.
As an AI Evangelist, I have always believed that ML would make a difference to the lives around me. At DSA Africa, I got to see first-hand how many lives we can change for the better. And we have only begun.
To find out more about what Arm and its ecosystem are doing in ML and Sustainability, visit our AI and sustainability pages on Arm.com.
To hear more from Arm Innovator and DSA founder Ciira Maina, read his blog, 'From capacity building to deployed solutions - The “end-to-end” Data Science Africa approach'.
Read blog