I would like to show a great appreciation to MBB grant for sponsoring this project. For me, it is a great gesture of believe and support for innovators in Africa. We live in a world where it is hard to be trusted to build anything of substance without too much bureaucracy and noise. After the award of the grant (#700,000) in December 2024, I kickstarted the project. The project was divided into 4 key parts which are the Overview Research, Prototype Development, AI Model Development and Project Testing.
Each of these processes is a series of back-and-forth interactions that served as a critical milestone for the project and taught me some great lessons. The project research involves one-on-one meetings, sharing questionnaires, and talking to potential users about the idea, both physically and on social media. The feedback was great, and I got the validation that ojumi is a needed solution that people are willing to pay for. This process took about a month and really got me excited.
.jpg)
For a prototype, I developed an android app with the concept of identity at its core. I focused on data collection, security and data privacy as I built the app. The app has critical features with ready-made use cases for testing. I wanted it to be practical, functional, scalable and secure for each process. This process took about 5 months after which the first Ojumi user interface video was created.
After the android app, I needed to be able to use it on my system so I developed the Ojumi desktop app. I realized the hardware limitations of the android app and often times the need for me to have a localized interface to test features and see further integrations. Furthermore, off-the-shelf solutions are too expensive so I made the effort to build and test locally because of the huge cost implication of cloud services for the ai modules. I only test on cloud services when and where necessary. To truly verify an identity, you need a combination of an array of information in various formats (could be documents, live video, pictures and input text) which needs to be crunched for proper identity detection. Hence, I decided to setup individual pipelines to test and improve each of these aspects.
For the prototype, I focused on the data collection, storage and formatting, user onboarding and ease of use. I also created quick use cases that can provide values to the app users and document the process along the way. At the core, ojumi relies on two cores processes, prior enrollment and efficient matching. In our testing process, I realized the efficiency of identification is directly influenced by the efficiency and cost of detection based on past enrolled data points.
Ojumi is a great project and MBB grant have made it transform from just an idea to something that is tangible and currently under scaling.
Currently, Ojumi is undergoing testing. I am also working on more ways to improve the efficiency of the software and do more demos. Next will be more advance ai detection, cloud scalability and portability features.
For more of this, follow [at]ojumiai on social media.