Computer Vision for Creative Optimization

This project focuses on using computer vision methods based on deep learning for creative optimization in mobile advertising. The customer has been running a sizable number of ads, and each one has its own creatives. These designs were produced based on the designers' prior work and the requirements of the business. Because of this, it is impossible to gauge a creative's potential performance before running them and to predict how well they will do after. As a result, the creation of deep learning-based algorithms that extract elements from creative materials and relate them to the KPI parameters of the related campaigns was the primary aim of this project.

A/B TESTING

This was a project for helping an advertisement company measure the effectiveness of the ads they made for a client, It was also build in a way that the company could use the system for measuring future ads performance. The project was completed in teams. I was responsible for setting up the MLOps components and the ML-based A/B testing pipeline.

AFRICAN LANGUAGES SPEECH-TO-TEXT

This was a team effort to build a Speech to text engine that would be able to transcribe two African languages. These were Amharic and Swahili. We explored some deep-learning architectures that would give the best results. And at the end, we build a web app that a user could use to interact with our model. You can find a lot more details here.