Africa Youth Employment Clock
The Africa Youth Employment Clock is a Mastercard Foundation project executed by the World Data Lab (WDL) to track and predict employment trends in Africa, and help policy-makers to make data-sriven decisions. At WDL, I led the data engineering and data science work for Ghana and Nigeria with sub-national granularity, including data acquisition, modelling and visualization, as well as presentations at the international launch events.
Document Semantic Search
Used multilingual embeddings to build a RAG tool for running semantic searches across multiple documents in order to answer questions based on the said documents with compatibility for 100+ languages.
This project won 1st place in the #CohereAIHack
CYI IT Projects (Tech Lead)
CYI is a Christian youth development organization based in Accra, Ghana, with branches across Africa and Europe. I lead development on a number of WordPress projects and JavaScript projects linked by a central API built from the ground up.
COVID-19 Policy Effectiveness Analysis
Code repo for the paper: A global analysis of the effectiveness of policy responses to COVID-19. Published in Nature Scientific Reports
Smartcity Africa Sentiments
Code repo for the paper: Highlighting smart city mirages in public perceptions: A Twitter sentiment analysis of four African smart city projects Published in Cities
Website Safety Analyzer
Fine-tuned an image classifier for detecting explicit sexual coontent. The model is wrapped in a script that scrapes every image from a given website and assesses the safety of the site for young visitors.
Gradio Demo on HuggingFace Spaces
Multi-WordCloud Generator, A Research Tool
A utility for generating word clouds from multiple documents at a go, cutting down manual iteration time. The tool comes with integrated OCR text extraction using Pytesseract, allowing users to create word clouds from scanned PDF files.
UNet Foreground Image Segmentation
A TensorFlow project from 2020 using a modified UNet autoencoder architecture with MobileNetV2 weights to perform transfer learning for foreground image segmentation. The original usecase was polyp segmentation from colonoscopy images, but this was repurposed for foreground segmentation of everyday objects.
Publications
- Agyapon-Ntra, K. and McSharry, P., 2023. A global analysis of the effectiveness of policy responses to COVID-19. Sci Rep 13, 5629.
- Arku, R., Buttazzoni, A., Agyapon-Ntra, K. and Bandauko, E., 2022. Highlighting smart city mirages in public perceptions: A Twitter sentiment analysis of four African smart city projects. Cities, 130, p.103857.
Certifications
- TensorFlow Developer Certificate - TensorFlow.org
- Deep Learning Specialization - DeepLearning.AI
- TensorFlow Developer Specialization - DeepLearning.AI