Finalist category: BT Connected Society Award
BT Connected Society Award, Finalist, 2018
Using machine learning to match refugees to life opportunities
Rafiqi is an online platform leveraging artificial intelligence to match refugees to opportunities that will accelerate their integration.
The creator, an immigrant from the Middle East who moved to Europe 10 years ago and relocated to the UK only recently, strongly believes that this cutting-edge technology will bring newcomers closer to the better future they are desperately seeking.
Where existing initiatives focus on providing a well-defined single service to refugees (e.g. finding a job, learning a language etc.), Rafiqi allows users to discover and navigate the services themselves and to be matched in real-time and in a customized fashion to the services that better suit their profiles and needs.
Services include: jobs, job coaching programs, technical trainings, language/cultural trainings, access to mentors, accredited certifications and university degrees. These services can be provided online or onsite (or both) and can be refugee-focused or open to a wider audience.
The logic and design of the platform stems from the creator’s background in computer science and data science, and from her on-the-ground experience, spending last year in discussions with stakeholders and undertaking dozens of manual matching experiences. This on-the-ground experience made the creator aware of some challenges hindering the realization of refugee integration prospects including:
- lack of a single platform where newcomers can navigate the range of opportunities available to them, and where NGOs, universities, employers and volunteering mentors can access and filter refugee talent
- Newcomers’ unawareness of opportunities and of the right opportunities results in them being unemployed or being overqualified for what they are doing on the long run
- While there are significant efforts by governments and NGOs to match refugees to opportunities, scaling these efforts cannot be done manually given the number and diversity of both refugee profiles and available opportunities
On the backend side, the platform works as follows: First, opportunities are clustered using metrics such as category, theme, delivery mode and eligible footprint. The matching process then uses a technique called ‘decision trees’ to refine opportunities using the refugee’s data, such as location, work and education background, job readiness, digital literacy and language level. Eventually, the most appropriate opportunities are identified.
The service is very fast – the user inputs their data in under two minutes, is provided with a list of opportunities and can rate the outcome straight away. It is also customized – the use of machine learning means it can be scaled up without losing the benefits of the individual matching process. As a single platform, users no longer need to go to multiple agencies or rely on their personal network – instead browsing the wide catalog of opportunities in one place, with NGOs, companies, universities and mentors also able to access and filter refugee talent.
Currently, a Minimum Viable Product (MVP) of the matching tool is available, with over 150 opportunities in the database. This will be added to further. The product is currently being piloted among a selected group of refugees in London and Berlin. Feedback will be used to fine-tune the interface and auto-adapt the machine learning algorithm. Once enough feedback has been gathered, the final version of the web app will be launched in the UK, where over 100000 refugees currently live, along with the development of a mobile app and an interactive chatbot that provides the same service and incorporates all the learnings.