Finding the right AI talent can be difficult – the right skills are hard to come by and are often very expensive. Dutch entrepreneur Pieter Boon found a solution to this challenge in Cape Town, South Africa. Through offshoring, Cape AI is able to provide their customers with highly skilled AI talent at a fraction of the normal cost.
Cape AI currently has consultants working for both South African and Dutch companies. Many of these client engagements are fully remote. In just two years, Cape AI has become a leading consulting firm in machine learning and data science solutions. As a result of their rapid growth and cost-effective strategy, Pieter and his team are now rapidly scaling their services across the European continent.
Pieter, how did you end up in South Africa?
‘When I left my previous company in the Netherlands, I felt it was time for a change. South Africa always appealed to me because of the beautiful weather, the high proficiency in English and the strong work ethos. In addition, because of the small time zone difference, it is very easy to work remotely for Dutch and European companies. My programming skills are limited though so I needed to find and work with highly skilled technical individuals. Fortunately, South Africa is home to great universities and ambitious people who want to make a difference in the country and do big things through technology and entrepreneurship. As the company grew, we felt we had the expertise to take on new challenges. This led to us starting the development of several ‘ventures’ – AI products which we develop in-house with the intention of creating social and commercial value. To date, we have had several ventures spin-out of Cape AI (Enlabeler, Knowledgemarker and Autoscriber), which are now all running independently as their own businesses. We are also very excited to be working on our latest venture – Moonshop – a grab-and-go autonomous shopping experience.’
What sets Cape AI apart from other consultancy firms?
‘We are able to maintain a lean cost-structure, which is primarily driven by lower costs of living in South Africa relative to Europe. This allows us to price our AI consulting services at more affordable rates than our counterparts, whilst still providing quality that is on par or higher than our European (or American) competitors. In addition, all of our European business activities go through our Dutch BV, so our clients have no trouble with foreign invoices or payment issues. In a South African context, we are also one of the few consultancy firms that specialise in tabular data, computer vision, natural language processing and MLOps/DevOps – offering a full spectrum of AI solutions.’
‘Another thing that sets us apart is that everyone in Cape AI gets to spend 20% of each week on technical training and personal development. This is core to our company values and is also really important to our employees. We are not only able to do technical consultancy, but we are also builders – technologists at heart! That sets us apart from other consultants who just sell hours.’ Lastly, we are able to create advanced and scalable algorithms and APIs, all with the average age of our company being under 28!’
What is the mission of Cape AI?
‘Our mission is to effect positive global change on the triple bottom line – People, Planet, and Profit – both for ourselves and our clients through data centric technologies. This mission is supported through 4 core company values: Trust, Excellence, Entrepreneurship and Fun. We take on all projects with the intention and focus to generate commercial value for clients as well as social value for our planet and society at large. This guides our focus in all domains from the projects we take on to the technical and personal upskilling we do internally.’
‘We recently started an ‘AI for Good’ initiative at Cape AI. This initiative is centred around using AI technologies to tackle socially and environmentally pertinent issues. The initiative is made up of a series of projects that achieve a balance between a high societal or environmental impact, and useful upskilling opportunities for Cape AI employees and other project volunteers. For these projects, we define very specific deliverables like a model or an API – always something very concrete. This allows our employees to get hands-on technical experience and growth in a low pressure consulting setting and do good for the planet and society at the same time.’
‘An example of our commitment to AI for Good is our participation in the ‘AI for Wildlife’ challenge, where computer vision models are used to spot poachers in wildlife parks. These on-edge technologies running on drones, created in partnership with a Dutch-based company FruitPunch AI, allow rangers to intercept poachers more efficiently.’
Cape AI has done work for several organisations spanning multiple industries. This spectrum ranges from medical institutions like LUMC to Webshops such as InstallatieBallie. One of the projects which we completed involved creating a data and AI solution for Opt Out Advertising – a technology partner for publishers, agencies and advertisers in the transition to cookieless advertising. The world of advertising is changing rapidly with demand for privacy-friendly solutions, whilst harnessing the data to track which campaigns are most effective. Cape AI ML Engineers Christo Rademan, Samuel Sendzul and Jeanne Danniel worked on adding to and improving the automation tools that Opt Out uses to help their clients serve advertisement content. Linda Worp, Business Development Manager at Opt Out Advertising, was the project lead from the client’s side.
Linda, why did you choose Cape AI as your AI-partner?
‘Privacy is becoming more and more important and we’re heading towards a future without cookies and tracking. Since using very specific personal information for advertisement targeting won’t be possible anymore, we needed to switch to alternative targeting options to stay relevant.’
‘We started working with Cape AI in the summer of 2021. Someone else mentioned them as a good partner. We were very lucky to find them because we don’t have our own data scientists and it’s very hard to find data scientists if you’re not in the tech space. Also, they’re usually quite expensive. In Cape AI, we found a partner who could think strategically with us. We wanted them to look at how we process our text and articles. For instance, how can we make sure we select the right context, subject, sentiment, emotion and keywords for the adverts we serve? Cape AI looked into what would be good for us.’ ‘They don’t just do what you tell them to do but instead challenge you when necessary to ensure you arrive at the best possible solution.
‘We needed someone to challenge our ways of thinking and decision-making. It was great to have someone else with a fresh perspective on our problem and what we are facing. They could think with us in order to find the right solutions. We worked with them on a few projects and we changed our methodology and focus during those projects thanks to their insights and guidance ’
Can you give an example of this change in focus?
‘At Opt Out, we are focused on changing the online environment. We acknowledge the fact that privacy is very important to the modern consumer and they should have the opportunity to opt out of having every detail of their internet browsing being tracked. That’s why we changed our focus from using personal information to using content driven information – information found on the website the consumer is viewing – to help us serve adverts to a website viewer. We changed our focus from the person behind the screen to what they are reading at this moment.’
How did Cape AI solve the problems you ran into?
‘We were already capable of predicting the context of website content, but we needed to make sure it was more accurate. So we partnered not only with Cape AI, but with Enlabeler – a company specialising in the labeling of unstructured data for machine learning. We sent thousands of articles from every genre to Enlabeler for data labeling and used these labeled datasets for training AI models. In the beginning, it was only specific news. After that, we added more diverse articles. That way we made sure the models work for all kinds of publishers and not just for specific news publishers. Cape AI and Enlabeler labeled everything with the right context. Also, the articles were read by multiple people so there was no bias or conclusions made based on a single opinion.’
Pieter: ‘Enlabeler was our first venture and it’s now run by Esther Hoogstad (another Dutchie). We started Enlabeler to meet the demand for high quality labeled datasets for machine learning modelling while also creating jobs in South Africa to help address youth unemployment. Enlabeler supports us and our customers with data labeling, annotations, and data quality assurance.’
Linda: ‘We needed to make sure we had a sizable data set in order to have a training, validation and test set. With all those labels, Cape AI was able to make predictions about the context and add a positive or negative sentiment classification to articles. We also weren’t able to predict the emotions present in articles at the time, but we are now able to predict emotion accurately thanks to the work done by Cape AI and Enlabeler. In our opinion, this is a really good add-on. Advertisers can now change the tone of voice in their advertisements based on the emotions and sentiment detected in articles. We have also added a brand safety model to predict whether an article is safe for a brand to advertise alongside (i.e. it doesn’t contain content that could damage the brand if associated with it).’
Christo: ‘It’s wonderful that we could help Linda and that she is happy with the solutions we have provided. Their problem space was quite unique. What we built, in a nutshell, were models which help her to identify what kinds of advertisements should be shown on their websites. We do that by analyzing the text on the websites using natural language processing techniques. That gives Linda the ability to automatically suggest which ad a publisher should put on a website. It was a very dynamic process, and quite fun as we were able to work with Linda to select models based on her long and short term strategies and feedback. We also worked to hand over models as individual components to help her realise value immediately.’
Finally, Pieter, why did you decide to partner up with AI.nl?
‘First of all, I really like Remy. We have great chemistry and I really believe in what he stands for. Also, we hope to attract more clients like Linda who know that hiring AI-talent can be tough and expensive. We want to let people know that there are other possibilities, like Cape AI, where you can hire a team for a complex project, for instance, but do this by offshoring. For us, being part of AI.nl is a great opportunity to get leads, grow our AI network, and ultimately contribute towards our company mission. I am convinced that they do that really well.
Cape AI is a contributing member to the ai.nl community – sharing knowledge, network and insights about data science, machine learning and artificial intelligence. If you want to know more about Cape AI and their value for end customers. Visit their company profile for more information.
