7 completed projects spanning financial technology, machine learning, NLP, healthcare, and community data systems. Each project is documented with methodology, results metrics, and downloadable code where available.
Production-grade AI-driven ALM intelligence platform at Technocolabs Softwares. Identified a severe maturity mismatch across 5 asset and liability buckets from a 10,000-row synthetic NBFC dataset.
XGBoost, Random Forest, and Logistic Regression on 50,000 transactions. ROC-AUC 0.983, PR-AUC 0.871. SMOTE oversampling, class imbalance handling, feature importance analysis.
TF-IDF with bigrams and Linear SVM on 5,572 SMS messages. 98.9% accuracy, 0.981 5-fold CV F1. PorterStemmer, stopword removal, discriminative word analysis.
scikit-learn Pipeline with ColumnTransformer eliminating data leakage. GridSearchCV-tuned Gradient Boosting on 7,043 telecom records. ROC-AUC 0.934, PR-AUC 0.812. Business impact analysis.
TF-IDF vectorisation with 15,000 bigram features on the 20 Newsgroups dataset (6 categories). Linear SVM: 94% accuracy and 0.94 weighted F1. Keyword analysis reveals discriminative terms per class.
Python data pipeline and choropleth geo-visualisations for the 2026 Kwara State Mass TADs Preliminary Report. Coverage analytics and Power BI dashboards for 16 local government areas.
Multi-sheet Excel workbook tracking 200+ households, payment history, arrears, and automated summary reports. Built for the Ifeloju Community security management committee in Ibadan.
Four of these projects have fully documented Jupyter notebooks available for direct download as .ipynb files.